Mapping CpG Sites to Quantify Aging Traits

CLAIM OF PRIORITY

This application claims the benefit of U.S. Provisional Application Serial No. 63/371,877, filed on August 19, 2022. The entire contents of the foregoing are incorporated herein by reference.

FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

This invention was made with Government support under Grant No. AG065403 awarded by the National Institutes of Health. The Government has certain rights in the invention.

TECHNICAL FIELD

Provided herein are methods that use methylation of causal CpG sites to quantify aging and predict whether an intervention will be protective or damaging to the aging process.

BACKGROUND

Aging is a complex biological process characterized by a buildup of deleterious molecular changes that result in a gradual decline of function of various organs and systems and ultimately lead to death k Although the underlying mechanisms of aging are not well understood, various studies indicate that aging is strongly associated with changes in the epigenome, quantified as a set of chemical modifications to DNA and histones that affect gene expression and chromatin structure ² . DNA methylation is one of the best studied epigenetic modifications. In mammals, 5-methylcytosine (5mC) is the most common form of DNA methylation, which is achieved by the action of DNA methyltransferases (DNMTs) ³ ’ ⁴ . Studies have shown that DNA methylation patterns change with age, wherein the global level of DNA methylation decreases slightly during adulthood, while some local areas may be hypomethylated or hypermethylated ^2,5 “ ⁹ . Furthermore, the level of methylation of some specific CpG sites shows a strong correlation with age, which can be used to build machine learning-based models that can accurately predict the age of biological samples ^{8 10} . As models can quantify age with very high accuracy, researchers termed these models epigenetic aging clocks (e.g., Horvath pan tissue epigenetic clock and Hannum blood based epigenetic clock) ^{11 12} . The predicted age based on various epigenetic aging clocks appears to have a higher association with health-related measurements than chronological age ^{13 14} . Therefore, it is believed that they could be used to better represent the biological age of samples than chronological age ¹⁵ .

SUMMARY

Provided herein are methods comprising providing a biological test system, optionally a cell, tissue, organ, or organism; and determining methylation of one or more causal CpG sites identified in Tables A, B, and/or C. As used herein, determining can include performing an assay (or causing an assay to be performed) on a test system, or can include using existing methylation data.

In some embodiments, the methods include determining methylation of 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 125, 150, or more of the causal CpG sites. In some embodiments, the methods include determining methylation of up to 100, 125, 150, 175, 200, 250, 300, 350, 400, 500, or 1,000 CpG sites, including determining methylation of at least 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 125, 150, or all of the causal CpG sites.

In some embodiments, the methods also include applying an intervention to the system and determining methylation of the one or more causal CpG sites during and/or after an application of an intervention.

In some embodiments, the methods further include comparing the methylation of the one or more causal CpG sites to a reference pattern of methylation. In some embodiments, the methods include the reference pattern is a baseline methylation pattern obtained in the same test system before application of an intervention, or a level obtained earlier in time in the same test system, or a level or range in a reference system that represents a level or range of methylation in the absence of an intervention.

In some embodiments, the methods include determining methylation of a plurality of causal CpG sites, and calculating a score using an algorithm to calculate the cumulative effect on aging or aging-related outcomes, optionally wherein the algorithm comprises using a manual or software-based modeling algorithm, optionally wherein the algorithm comprises a linear algorithm; principal component analysis (PCA); classification or decision trees; elastic net analysis; linear and polynomial support vector machines (SMV); shrunken centroids; random forest algorithms; support vector machines; or neural networks.

In some embodiments, the methods include calculating a predicted age using the determined methylation and applying an algorithm to the levels.

In some embodiments, the algorithm comprises :

PredictedAge = intercept + bl * CpGl + b2 * CpG2 +...+ bn * CpGn

Where bl - bn are the model coefficient ‘estimate’ from Tables A, B, and/or C and CpGl - CpGn are the methylation level of given CpG sites (on a scale of 0-1).

In some embodiments, the methods include identifying an intervention as having a protective effect when changes in methylation are observed that are consistent with protection, and/or identifying an intervention as having a damaging effect when changes in methylation are observed that are consistent with damage.

In some embodiments, the methods also include: selecting an intervention that has been identified as having a protective effect as a candidate intervention; applying the candidate intervention to an in vivo model of a disorder or condition associated with aging, optionally wherein the model is a non-human test animal or a human subject in a clinical trial; and determining whether the candidate intervention has a protective effect on the disorder or condition related to aging.

Also provided herein are methods of predicting an effect of an intervention on aging. The methods include: providing a biological test system, optionally a cell, tissue, organ, or organism; and determining methylation of one or more causal CpG sites identified in Tables A, B, and/or C; applying an intervention to the system, determining methylation of the one or more causal CpG sites during and/or after an application of an intervention; comparing the methylation of the one or more causal CpG sites to a reference pattern of methylation; and identifying an intervention as having a protective effect on aging when changes in methylation are observed that are consistent with protection, and/or identifying an intervention as having a damaging effect on aging when changes in methylation are observed that are consistent with damage.

In some embodiments, the methods include determining methylation of 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 125, 150, or more of the causal CpG sites.

In some embodiments, the methods include determining methylation of up to 100, 125, 150, 175, 200, 250, 300, 350, 400, 500, or 1,000 CpG sites, including determining methylation of at least 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 125, 150, or all of the causal CpG sites.

In some embodiments, the reference pattern is a baseline methylation pattern obtained in the same test system before application of an intervention, or a level obtained earlier in time in the same test system, or a level in a reference system that represents the level of methylation in the absence of an intervention.

In some embodiments, the methods include determining methylation of a plurality of causal CpG sites, and calculating a score using an algorithm to calculate the cumulative effect on aging or age-related outcomes, optionally wherein the algorithm comprises using a manual or software-based modeling algorithm, optionally wherein the algorithm comprises a linear algorithm; principal component analysis (PCA); classification or decision trees; elastic net analysis; linear and polynomial support vector machines (SMV); shrunken centroids; random forest algorithms; support vector machines; or neural networks.

In some embodiments, the methods include calculating a predicted age using the determined methylation and applying an algorithm to the levels.

In some embodiments, the algorithm comprises :

PredictedAge = intercept + bl * CpGl + b2 * CpG2 +...+ bn * CpGn where bl - bn are the model coefficient ‘estimate’ from Tables A, B, and/or C and CpGl - CpGn are the methylation level of given CpG sites (on a scale of 0-1).

In some embodiments, the methods include selecting an intervention that has been identified as having a protective effect as a candidate intervention; applying the candidate intervention to an in vivo model of a disorder or condition associated with aging, optionally wherein the model is a non-human test animal or a human subject in a clinical trial; and determining whether the candidate intervention has a protective effect on the disorder or condition related to aging.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Methods and materials are described herein for use in the present invention; other, suitable methods and materials known in the art can also be used. The materials, methods, and examples are illustrative only and not intended to be limiting. All publications, patent applications, patents, sequences, database entries, and other references mentioned herein are incorporated by reference in their entirety. In case of conflict, the present specification, including definitions, will control.

Other features and advantages of the invention will be apparent from the following detailed description and figures, and from the claims.

DESCRIPTION OF DRAWINGS

FIGs. 1A-E. Epigenome-wide Mendelian Randomization on various aging-related phenotypes, a. Schematic diagram shows the principle of MR using meQTLs as exposures and aging-related traits as outcomes to identify putative causal CpG sites, b. Flow chart shows the procedure for epigenome-wide MR and sensitivity analysis, c. Number of significant putative causal CpG sites identified for each trait after adjusting for multiple tests using the Bonferroni correction. Red regions of the bars indicate the number of putative causal CpG sites supported by the colocalization analysis with conditional PP-H4 > 0.7. d. Spearman correlation of the estimated causal effects of CpGs in twelve traits. Only CpGs with significant MR signals across at least six traits are included in the analysis. Color scheme reflects Spearman correlation coefficients, * adjusted P < 0.05, ** adjusted P < 0.01, *** adjusted P < 0.001. e. Modified Mississippi plot shows significant MR signals for Aging-GIPl. X- axis corresponds to the genomic positions of CpG sites; Y-axis represents the size of the causal effect adjusted by colocalization probability (PP-H4). CpG sites with top adjusted causal effects are annotated with the name and nearest gene. Only CpG sites with adjusted P < 0.05 are included in the plot.

FIGs. 2A-F. CpG sites causal to aging are enriched in specific genetic regulatory regions, a. Bar plot shows enrichment of putative causal CpG sites in 14 Roadmap genomic annotations. Y axis shows -loglO (FDR) based on Fisher’s exact test, signed by log2 (Odds ratio). Putative causal CpG sites identified for different traits are annotated with different colors. Two dotted horizontal lines show the FDR threshold of 0.05. TssA, active transcription start site. Prom, upstream/downstream TSS promoter. Tx, actively transcribed state. TxWk, weak transcription. TxEn, transcribed and regulatory Prom/Enh. EnhA, active enhancer. EnhW, weak enhancer. DNase, primary DNase. ZNF/Rpts, state associated with zinc finger protein genes. Het, constitutive heterochromatin. PromP, Poised promoter. PromBiv, bivalent regulatory states. ReprPC, repressed polycomb states. Quies, quiescent state, b, c. Box plot shows distribution of conservation scores in causal and non-putative causal CpG sites for Aging-GIPl. Conservation scores are obtained by Learning Evidence of Conservation from Integrated Functional genomic annotations (LECIF, b) and phastCons (c). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. d. Enrichment of putative causal CpG sites for 12 aging-related traits against transcription-factor-binding sites. Each horizontal bar represents an enriched term. The X-axis shows the -logl0( -value), signed by log2 (Odds ratio). The top 10 enriched terms that passed the FDR threshold of 0.05 for each direction are annotated, e. Scatter plot showing the mediation analysis of Aging-GIPl. The total causal effects are shown on X-axis and the direct effects of DNA methylation are shown on Y-axis. The color shows the proportion of DNA methylation (DNAm) causal effect that is mediated by gene expression. Top CpG-gene pairs are annotated, f. Enrichment of the top mediator gene for Aging-GIPl in GO terms (above dashed line) and KEGG pathways (below the dashed line). The X-axis shows -log(P) for the fisher exact test.

FIGs. 3A-G. MR on epigenetic age successfully recovers clock sites as putative causal CpG sites, a. For epigenetic age measurements, true causal sites are the clock sites and the sites upstream of clock sites. We used these traits as a positive control to validate the MR approach, b. Forest plot shows enrichment of clock sites for each model in putative causal CpG sites. For each clock trait, putative CpG sites are identified by MR using corresponding clock traits as outcome. X-axis shows the log2(Odds Ratio). -values calculated by Fisher’s exact test are annotated. Error bars show 95% confidence intervals. Different colors represent different thresholds for putative causal CpGs. c-e. Correlation between ground truth causal effects (clock coefficients, X-axis) and causal effects estimated by MR (Y-axis, using GWAS of corresponding clocks as outcome traits) for Hannum age (c), Horvath age (d) and PhenoAge (e). Different colors represent different thresholds for putative causal CpGs. Pearson's correlation coefficients and -values are annotated, f. Receiver operating characteristic (ROC) curves show sensitivity (Y-axis) and 1 -specificity (X- axis) of MR in identifying putative causal CpG sites for clock traits, with the area under the ROC curve (AUC) annotated, g. Forest plot shows enrichment of clock sites for six aging clock models in putative causal CpG sites identified by MR for each trait. X-axis shows the log2(Odds Ratio). -values calculated by Fisher’s exact test are annotated if P < 0.05. Error bars show 95% confidence intervals. Different colors represent the different thresholds for putative causal CpGs. FIGs. 4A-C. Integration of causal information and age-associated differential methylation to separate protective and damaging epigenetic changes. a. Schematic diagram showing the method to identify protective and damaging epigenetic changes by integrating MR results and age-related differential methylation. b. Relationship between MR-estimated causal effects (X-axis) and age-related differential methylation (Y-axis) for each significant putative causal CpG identified in Aging-GIPl. The color scheme highlights the expected impact of age-related differential methylation on aging. Error bars show the standard error of b. The size reflects the PP-H4. Only CpG sites with adjusted -values < 0.05 and relative PP-H4 > 0.7 are plotted. The CpG sites with the top 10 largest effect sizes are annotated, c. Area plots show the total cumulative effect of changes in DNA methylation on Aging- GIP1. X-axis shows the rank of top 3,000 CpG sites based on the magnitude of age- associated differential methylation. Y-axis and the color scheme show the P-value estimated by 10,000 permutation tests.

FIGs. 5A-E. Construction and application of causality-informed epigenetic clocks, a. Schematic diagram shows the procedure of constructing causality-informed epigenetic clocks, b. Scatter plots show the accuracy of causal clocks on the test set. The X-axis shows the real age of each sample, and the Y-axis shows the predicted age of the same sample based on each clock model. Median absolute error (MAE) and Pearson’s R are annotated, c. Line plot showing the relationship between causality factor (T) and clock accuracy measured by MAE and Pearson’s R. d. Line plot shows the relationship between the causality factor (T) and - loglO(p) for the association with mortality risk (signed by log2(hazard ratio)) estimated from the meta-analysis of FHS and WHI cohorts. Yellow dashed line shows the P threshold of 0.05. Hazard ratio of mortality risk for every 10-year increase in age for each clock model and the 95% confidential interval for T = 0.3 is annotated. Results based on Horvath age, Hannum age, and PhenoAge are also shown by arrows for comparison, e. Scatter plots show the application of causal clocks and five other aging clocks to reprogramming of fibroblasts to iPSCs. X-axis shows days after initiating reprogramming. Pearson's R and P values are annotated.

FIGs. 6A-C. Causality-informed epigenetic clocks can better capture aging-related effects, a. Box plots show the association between epigenetic age and aging-related conditions, including atherosclerosis, prostate cancer prognosis, and hypertensive heart disease, b. Box plots show the association between epigenetic age and damaging conditions, including smoking, Progeroid Syndromes, and Sun exposure. Scatter plots show the correlation between epigenetic age and blood P0N1 activity. Epigenetic age prediction is rescaled to a 0-1 scale for better comparison. The color scheme shows the PON1 genotype in subjects. Linear regression is performed, and Pearson’s R and P values are annotated, c. Box plots show the association between epigenetic age and short-term treatments, including the umbilical cord blood plasma treatment, 15 months of cigarette smoke condensate (CSC) treatment, and 6- week supplementation of overweight subjects with omega-3 fatty acids. For umbilical cord blood plasma treatment, paired sign-test was performed, and the color scheme and the pie chart indicate whether the subject is rejuvenated after treatment based on the corresponding clock. For unpaired box plots, significant pairs based on two-tail t- test are annotated with stars. * P < 0.05, ** P < 0.01, *** P < 0.001, **** p < 0.0001.

FIG. 7. Genetic correlation between 12 lifespan-related phenotypes.

Genetic correlations were calculated using LDSC. Areas of the squares represent absolute values of corresponding genetic correlations. Genetic correlations that could not be estimated are shown as blanks. P values are corrected using Bonferroni correction for the number of tests, * P nominal < 0.05, ** P adjusted < 0.05, *** P adjusted < 0.01.

FIGs. 8A-B. Relationship between meSNPs and causal CpGs. Forest plot shows enrichment of meSNP among causal CpGs. Error bar shows the 95% confidential interval. P-value of significant results is annotated (8A). Scatter plot shows Pearson's correlation between the effect of a single CpG site estimated by MR and a single meSNP (8B). Correlation coefficient and P-value are annotated at the top.

FIG. 9. Relationship between estimated causal effects and evolutionary conservation. Box plot shows the distribution of conservation scores in causal and non-causal CpG sites. Conservation scores were obtained by Learning Evidence of Conservation from Integrated Functional genomic annotations (LECIF), phastCons, and phyloP. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

FIG. 10. Enrichment analysis. Bar plot shows enrichment of causal CpG sites in genomic annotations. Y-axis shows -logl0(FDR) based on Fisher's exact test, signed by log2(Odds Ratio). Causal CpG sites identified for different traits are annotated with different colors. Two dotted horizontal lines show the FDR threshold of 0.05. FIG. 11. Enrichment of causal CpG sites among CpG sites that show age- related changes. Error bar indicates the 95% confidence interval. Bar plot shows the signed -log 10(P -value) of Spearman's correlation between age-related change and causal effect size. The orange dotted line shows the threshold of P < 0.05.

DETAILED DESCRIPTION

Although epigenetic aging clocks provide a useful tool for profiling biological aging, they should be used with caution, as they are built based on pure correlations ¹⁶ . It is unclear whether differential DNA methylation used to predict age is causal to aging-related phenotypes or simply represents byproducts of the aging process that do not influence aging themselves. To establish a causal relationship, the gold standard approach is the application of randomized controlled trials (RCT), where participants are randomly assigned to the intervention arm that receives the treatment or the control arm. As the randomization step balances all confounding factors between two arms, the differences observed in the outcome between two groups are purely driven by the intervention; thus, the causal effect can be estimated ¹⁷ . However, given the large number of CpG sites across the genome, it is inefficient and infeasible to perform the perturbation on each of them and assess the aging-related outcomes.

Mendelian randomization (MR) is a genetic approach to causal inference that recapitulates the principle of RCT. Instead of perturbing an exposure through treatment, the MR uses the genetic variants that are robustly associated with the exposure as instrumental variables ^{18 19} . As genetic variants of parental DNA are naturally randomly passed on to the offspring, the effect estimated by MR is not affected by environmental confounders and thus can be considered as an estimation of a causal effect, similar to the RCTs. In recent years, several studies have shown that MR can be applied to molecular traits by using the genetic variants associated with molecular levels as instruments (also known as molecular quantitative trait loci, molQTL) ²⁰ . These molecular QTLs include gene expression (eQTL) ²¹ , RNA splicing (sQTL) ²² , plasma protein (pQTL) ²³ , metabolites (mQTL) ²⁴ , as well as DNA methylation (meQTL) ²⁵ . A previous study showed that it is feasible to use meQTLs as instruments to identify putative causal CpG sites for diseases ²⁶ . By integrating molQTLs with genome-wide association studies for traits such as lifespan, healthspan, extreme longevity, and other measurements related to aging, it is biologically plausible to perform two-sample MR to estimate the causal effects of molecular changes on the aging process.

Here, we leveraged large-scale genetic data and performed epigenome-wide Mendelian Randomization (EWMR) on 420,509 CpG sites to identify CpG sites that are causal to twelve aging-related traits. We found that none of the existing clocks are enriched for putative causal CpG sites. We further constructed a causality -informed clock based on this inferred causal knowledge, as well as clocks that separately measure damaging and protective changes. Their applications provide direct insights into the aging process. Thus, our results offer a comprehensive map of human CpG sites causal to aging traits, which can be used to build causal biomarkers of aging and assess novel anti-aging interventions and aging-accelerating events.

Many existing epigenetic aging clock models accurately predict the age of samples ⁸ , and there are numerous CpG sites that are differentially methylated during aging ²⁷ . DNA methylation levels affect the structure of chromatin and the expression of neighboring genes ^28,29 , through which they can causally affect aging-related phenotypes. A recent study also suggested that DNA methylation may play a causal role in the rejuvenation effect observed during iPSC reprogramming ³⁰ . However, it is important to understand whether age-related differential DNA methylation causes aging-related phenotypes and which of its components do it. A previous transcriptome-wide MR study revealed that differentially expressed genes in human diseases mainly reflect gene expression caused by disease rather than disease-causing genes ³¹ . Similarly, differential DNA methylations during aging may primarily reflect the downstream effects of aging phenotypes rather than causing them. Our EWMR findings support this notion as we found no significant overlap between CpG sites causal to healthy longevity and those differentially methylated during aging.

MR is a powerful method to identify causal relationships between exposure traits and phenotypes ³² . However, it is limited by the availability of genetic instruments for the exposure traits. In our study, we utilized the DNA meQTLs of 420,509 CpG sites from the Illumina 450K methylation array as instrumental variables to infer their causal relationship with aging-related phenotypes. However, there are many unmeasured CpG sites across the genome, and the methylation patterns of nearby CpG sites are highly correlated ²⁸ . Therefore, it is not possible to fully separate the causal effect of a single CpG and its neighbors. Analysis of point mutations at putative causal CpG sites (meSNPs) suggests that the epimutation of a single causal CpG site identified by MR may be sufficient to alter the phenotype (FIGs. 8A-B). However, due to the lack of abundance of meSNPs on putative causal CpG sites, this hypothesis is difficult to test across all causal CpG sites identified. Therefore, we tend to reach a more conservative conclusion and think that the putative causal CpG sites identified in our study serve as tagging CpG sites for causal regulatory regions in aging-related phenotypes. Future genome-wide meQTL studies may facilitate further analyses of causal effects of CpG sites at base-pair resolution.

The genetic instruments of CpG sites for our study were selected from the currently largest meQTL study in whole blood (GoDMC, 36 cohorts, including 27,750 European subjects). Therefore, the CpG sites we identified are valid in blood. However, a previous study showed that up to 73% cis-meQTLs are shared across tissues (including blood, brain, and saliva) ³³ . This suggests that the identified putative causal CpG sites also act in other tissues to affect lifespan and healthspan.

We found that TF-binding sites of BRD4 and CREB1 are enriched with CpG sites whose methylation levels promote healthy longevity, and TF-binding sites for HDAC1 are enriched with CpG sites whose methylation levels decrease healthy longevity. BRD4 is known to contribute to cell senescence and promote inflammation ³⁴ . Therefore, our findings suggest that higher DNA methylation at BRD4 binding sites may inhibit the downstream effects of BRD4 and promote healthy longevity. Similarly, previous studies showed that CREB1 is related to type II diabetes and neurodegeneration ³⁵ and mediates the effect of calorie restriction ³⁶ . However, how DNA methylation may affect CREB 1 binding is not well understood. Our data suggest that higher methylation at CREB 1 -binding sites may support its longevity effects. HDAC1 is a histone deacetylase, and its activity increases with aging and may promote age-related phenotypes ^30,37 . HDAC1 has been shown to specifically bind to methylated sites. Our data, therefore, support the hypothesis that HDAC1 plays a damaging role during aging, as increased DNA methylation a HDACl binding sites may causally inhibit healthy longevity.

One general approach for developing anti-aging interventions is to identify molecular changes during aging and use these changes as targets to modulate the aging process ^38,39 . A similar idea has also been applied to evaluate potential longevity interventions. However, this logic is intrinsically flawed, as correlation does not imply causation and age-associated differential methylation are not necessarily causal to age-associated declines. As living organisms are complex systems with various adaptive mechanisms, many molecular changes during aging are potentially neutral downstream effects of fundamental damaging changes or even adaptive mechanisms that protect against aging phenotypes. This notion is usually underappreciated as age- associated differential methylation are assumed to be damaging. As a result, adaptive mechanisms of aging are largely understudied. However, there is evidence to suggest that at least some age-associated differential methylation is protective against aging phenotypes.

An example of age-related protective changes is the Insulin and IGF-1 signaling (IIS) pathway. Attenuation of IIS signaling intensity through multiple genetic manipulations has been shown to consistently extend the lifespan of worms, flies, mice, and potentially humans ^40,41 . This pathway also mediates pro-longevity effects of dietary restriction ⁴⁰ . Growth hormone is produced by the anterior pituitary gland and can induce the production of IGF-1, thus increasing IIS signaling. Both growth hormone and IGF-1 levels decline during aging ⁴² , which is considered to be a defensive response that extends lifespan ⁷ . Another example of an age-related adaptation is protein aggregation. It has been shown in C. elegans that the protein aggregation events are increased during aging. Although it may look like a result of losing proteostasis, it turns out to be a protective mechanism that drives aberrant proteins into insoluble aggregates to improve overall proteostasis, and has been observed in long-lived mutants ⁴³ . Similar protective mechanisms are also observed in mouse nerves at the transcriptomic level ⁴⁴ .

The present results suggest that adaptive mechanisms at the epigenetic level are nearly as common as damaging changes and that simply following age-associated differential methylation in DNA methylation does not allow us to infer positive, neutral, or negative effects on age-related traits. However, the identified damaging and protective CpG sites are extremely useful both for understanding aging and quantifying it, and the same applies to rejuvenation. Together, the identified CpGs represent causal epigenetic changes, and their combined effect on health-related phenotypes is negative.

The framework we described for epigenetic changes in this study may be applied to any other age-related change, e.g., changes in the transcriptome, metabolome, and proteome. While all age-related features may be used to construct aging clocks, some of them are expected to be negative, some neutral, and some protective. Neither the direction nor the degree of age-associated differential methylation is important, and inferring the need to bring these changes to those observed in the young state as a way to rejuvenate an organism is equally incorrect. Instead, the focus should be on the causal effects of age-associated differential methylation, as well as on the direction of their effect. The present causal analysis was conducted using blood samples because large meQTL studies are only available in blood up-to-date ⁴⁵ . However, previous studies suggest that the cis-meQTLs are conserved across tissues ³³ , therefore the present findings are also likely applicable to other tissues.

The causal epigenetic clock models, CausAge, AdaptAge, and DamAge, could help separate protective changes from damaging events. We also showed that by preselecting the CpG sites that show protective adaptation during aging, it is possible to build an aging clock showing an inverse relationship with mortality. Specifically, subjects with elevated protective adaptation are predicted to be age-accelerated by AdaptAge and have a lower risk of mortality (FIG. 5c). Similarly, AdaptAge shows an inverse relationship with rejuvenation (e.g., iPSC reprogramming) and aging acceleration. Note that both DamAge and AdaptAge show similar accuracy in predicting chronological age, but their delta-age term reflects an opposite biological meaning. Although we observed a weak positive correlation between DamAge and AdaptAge in the general population, this correlation may be due to collider bias and survival bias ^46,47 , e.g., both DamAge and AdaptAge contribute to mortality and the individuals with high DamAge and low AdaptAge are removed from the population due to higher mortality risk, thus resulting in an apparent positive correlation. The causality-informed clock models described herein provide novel insights into the mechanisms of aging and provide methods for testing interventions to delay aging and reverse biological age.

Test Systems

Thus, provided herein are methods for identifying compounds or conditions that can be used to monitor effects of various interventions on methylation of CpG sites that affect aging, and to identify interventions that can delay or reverse the aging process in a tissue or a subject.

The methods can be practiced using a biological test system, including one or more human cells, all or part of a human tissue, or all or part of an human organ. The cell can be, e.g., a mammalian cell, such as a primary cell (including erythrocytes; platelets; peripheral blood mononuclear cells (PBMC), e.g., lymphocytes, monocytes, or macrophages; bone marrow cells; endothelial cells, e.g., vascular or bronchial endothelial cells; pancreatic islet beta cells; renal cells; hepatocytes; neurons and glia; epidermal cells; respiratory interstitial cells; adipocytes; dermal fibroblasts; muscle cells; cells of the eye (e.g., photoreceptors, RPE cells, retinal ganglia cells) or ear (e.g., hair cells or supporting cells); or hair follicles. Primary or cultured cells including stem cells and immortalized cells can also be used, e.g., induced pluripotent stem cells (iPSCs), embryonic stem cells (ES cells), hematopoietic stem cells (HSCs), mesenchymal stem cells (MSCs), pre-adipocytes, and neural progenitor cells.

Cultured cells such as HEK293 and fibroblasts can also be used.

The tissues can be, e.g., connective tissue, epithelial tissue, muscle tissue, and nervous tissue. The organs can be, e.g., capillaries; joints; nerves; skin; tendons; arteries; cerebellum; liver; nasal cavity; spleen; tongue; appendix; diaphragm; lungs; ovaries; scrotum; thyroid; adrenal glands; ears; larynx; esophagus; stomach; trachea; brain; eyes; ligaments; penis; spinal cord; thymus gland; bones; fallopian tubes; lymph nodes; pancreas; small intestine; ureters; bronchi; genitals; large intestine; pharynx; salivary glands; urethra; bladder; gallbladder; lymphatic vessel; placenta; skeletal muscles; uterus; bone marrow; heart; mouth; prostate; seminal vesicles; vulva; bulbourethral glands; hair follicle; mesentery; pineal gland; subcutaneous tissue; veins; colon; hypothalamus; mammary glands; pituitary gland; teeth; vagina; cervix; interstitium; nose; parathyroid glands; tonsils; vas deferens; clitoris; kidneys; nails; anus; rectum; or testes.

In some embodiments, the biological test system is whole blood, or a cell from an embryo, e.g., a human embryo.

In some embodiments, a whole organism is used; the organism can be, e.g., a human, optionally a human subject in a clinical trial or a veterinary subject in a clinical trial, or a non-human model animal, e.g., a non-human mammal such as a mouse, rat, or rabbit, or can be a nematode, insect (e.g., drosophila), yeast, or bacterium.

Interventions

The present methods can include applying one or more interventions to the test system. Interventions can include, for example, administration of one or more compounds, e.g., polypeptides, polynucleotides, or inorganic or organic large or small molecule test compounds. The intervention can also be, e.g., alteration of an environmental factor, e.g., food (e.g., quality or quantity of nutrition, calories, or type); exposure to toxic or potentially toxic environments (e.g., to mimic exposure to pollution or smoking); oxygen levels; and so on. When more than one intervention is applied, the more than one can include multiple applications over time of the same intervention, or application of multiple interventions, e.g., at the same time or consecutively or over time.

As used herein, “small molecules” refers to small organic or inorganic molecules of molecular weight below about 3,000 Daltons. In general, small molecules useful for the invention have a molecular weight of less than 3,000 Daltons (Da). The small molecules can be, e.g., from at least about 100 Da to about 3,000 Da (e.g., between about 100 to about 3,000 Da, about 100 to about 2500 Da, about 100 to about 2,000 Da, about 100 to about 1,750 Da, about 100 to about 1,500 Da, about 100 to about 1,250 Da, about 100 to about 1,000 Da, about 100 to about 750 Da, about 100 to about 500 Da, about 200 to about 1500, about 500 to about 1000, about 300 to about 1000 Da, or about 100 to about 250 Da).

The test compounds can be, e.g., natural products or members of a combinatorial chemistry library. A set of diverse molecules should be used to cover a variety of functions such as charge, aromaticity, hydrogen bonding, flexibility, size, length of side chain, hydrophobicity, and rigidity. Combinatorial techniques suitable for synthesizing small molecules are known in the art, e.g., as exemplified by Obrecht and Villalgordo, Solid-Supported Combinatorial and Parallel Synthesis of Small- Molecular-Weight Compound Libraries, Pergamon-Elsevier Science Limited (1998), and include those such as the “split and pool” or “parallel” synthesis techniques, solid-phase and solution-phase techniques, and encoding techniques (see, for example, Czarnik, Curr. Opin. Chem. Bio. 1 :60-6 (1997)). In addition, a number of small molecule libraries are commercially available. Natural compounds such as vitamins and neutraceuticals can also be tested using the present methods.

Methods for Determining Effects on Aging

The present methods include determining methylation of one or more causal CpG sites identified herein, i.e., in Tables A, B, and/or C. In some embodiments, the methods include determining methylation of 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 125, 150, or more causal CpG sites described herein; in some embodiments, the methods include determining methylation of up to 100, 125, 150, 175, 200, 250, 300, 350, 400, 500, or 1,000 CpGs, including at least 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 125, 150, or all of the causal CpG sites described herein. The methods can include applying an intervention to the system and determining methylation of the one or more CpG sites during and/or after application of the intervention.

As used herein, determining can include performing an assay (or causing an assay to be performed) on a test system, or can include using existing methylation data. Methods (assays) for determining methylation of a specific site are known in the art, and include sodium bisulfite conversion and sequencing (e.g., next-generation sequencing (NGS)), differential enzymatic cleavage ofDNA, CpG DNA methyltransferase, and affinity capture of methylated DNA; DNA affinity capture methods include methylated DNA immunoprecipitation (Me-DIP) that uses a methyl DNA specific antibody, or methyl capture using methyl-CpG binding domain (MBD) proteins. See, e.g., Tang et al., Methods Mol Biol. 2015;1238:653-75; Chatterjee et al., Methods Mol Biol. 2017;1537:249-277; Beck, Nat Biotechnol. 2010 Oct;28(10): 1026-8; Nair et al., Epigenetics. 2011 Jan;6(l):34-44; Hsu et al., Methods Mol Biol. 2020;2102:225-234; Feng and Lou, Methods Mol Biol. 2019; 1894: 181 - 227.

In some embodiments, the methods include comparing methylation of one or more causal CpG sites identified herein to a reference pattern of methylation. The reference pattern can be, e.g., a baseline obtained in the same test system, or a level or range obtained earlier in time in the same test system, or a level or range in a reference system that represents the level or range of methylation in the absence of an intervention. The reference system is typically the same type as the test system (i.e., a matched control) and be as identical to the test system as possible.

A test compound can be identified as having a protective effect when changes in methylation are observed that are consistent with protection as shown herein, i.e., reduce the age predicted by DamAge; conversely, a test compound can be identified as having a damaging effect when changes in methylation are observed that are consistent with damage as shown herein, i.e., increase the age predicted by DamAge. A change in methylation associated with a damaging effect will have the same directionality as shown in Table B, and a change in methylation associated with a protective effect will have the same directionality as shown in Table C. Where a plurality (more than one) level of methylation is determined, an algorithm can be used to calculate the cumulative effect on aging, e.g., manual or software-based modeling algorithms such as a linear algorithms, e.g., a rank-based linear algorithm; principal component analysis (PCA); classification or decision trees; elastic net analysis; linear and polynomial support vector machines (SMV); shrunken centroids; random forest algorithms; support vector machines; or neural networks.

For example, the methods can include calculating a predicted age using the determined levels of methylation and applying an algorithm to the levels. An exemplary algorithm is as follows:

PredictedAge = intercept + bl * CpGl + b2 * CpG2 +...+ bn * CpGn Where bl - bn are the model coefficient ‘estimate’ from Table A and CpGl - CpGn are the methylation level of given CpG sites (on a scale of 0-1, e.g., 0.7 means 70% methylated).

A similar algorithm can be used to quantify the age-related damage effect or protective effect of interventions using the model from tables B and C, respectively.

In some embodiments, the methods include summing the product of methylation difference and causal effect estimate for each cPG site, and determining if the sum is positive (i.e., more adaptation, thus protective) or negative (i.e., more damage, thus damaging). In some embodiments, the difference of the predicted age before and after treatment is used. For example, DamAge measures age-related damage, and if it is increased means that there are damage accumulated (usually bad). AdaptAge measure age-related adaptation/protection, if it is increased means that the adaptation is increased (usually good but can be bad or neutral)

An intervention that has been screened by a method described herein and determined to have a protective effect on aging can be considered a candidate compound. A candidate compound that has been screened, e.g., in an in vivo model of a disorder such as a non-human test animal or a human subject in a clinical trial, and determined to have a protective effect and/or a desirable effect on aging, e.g., on one or more symptoms of aging, can be considered a candidate therapeutic agent. Candidate therapeutic agents, once screened in a clinical setting, are therapeutic agents. Candidate compounds, candidate therapeutic agents, and therapeutic agents can be optionally optimized and/or derivatized, and formulated with physiologically acceptable excipients to form pharmaceutical compositions. The methods can also be used to identify interventions that are damaging, e.g., that can speed aging or cause premature aging; such interventions can be identified for avoidance or exclusion, e.g., in food, cosmetics, or pharmaceuticals.

Test compounds identified as protective hits can be considered candidate therapeutic compounds, useful in slowing, delaying, or even reversing aging. A variety of techniques useful for determining the structures of “hits” can be used in the methods described herein, e.g., NMR, mass spectrometry, gas chromatography equipped with electron capture detectors, fluorescence and absorption spectroscopy. Thus, the invention also includes compounds identified as “hits” by the methods described herein, and methods for their administration and use in the treatment, prevention, or delay of development or progression of a disorder described herein.

Test interventions identified as candidate protective interventions compounds can be further screened by administration to a test system in an animal model of aging, e.g., as described herein. The animal can be monitored for a change in aging, e.g., for an improvement in a parameter of aging, e.g., a parameter related to health or clinical outcome. In some embodiments, the parameter is development of age-related conditions such as hearing loss, cataracts and refractive errors, back and neck pain and osteoarthritis, chronic obstructive pulmonary disease, diabetes, and dementia, and an improvement would be a delay or decrease in risk of development of one or more age-related conditions. In some embodiments, the test system is epidermis, and the parameter is development of age-related skin conditions such as thinning, sagging, wrinkling, xerosis, pruritis, eczematic dermatitis, purpura, and chronic venous insufficiency, and an improvement would be a delay or decrease in risk of development of one or more age-related conditions.

Table A - CausAge CpG sites and estimates term estimate term estimate (Intercept) 86.8081638 cg00715290 -10.219189 cg00027162 1.66785269 cg00879155 0.61301692 cg00048759 5.41958522 cg00910168 -1.1934856 cg00200653 -0.26977 cg00962755 1.0630155 cg00347863 4.10387211 cg01035616 1.81673898 cg00505045 12.0066436 cg01048752 -1.1464102 cg00563845 -0.54509 cg01105058 5.10947032 cg00603274 0.1468829 cg01274524 0.29682433 cg00614360 1.188062 cg01321673 0.84088792 cg00655552 -0.9658713 cg01329511 3.69864743 cg00663739 3.54837844 cg01334432 -1.8983001 term estimate term estimate cg01399860 -0.3860917 cg03438101 2.0009373 cg01421252 -0.8981966 cg03446427 -0.3279279 cg01454752 5.88142509 cg03520471 -1.506224 cg01503516 1.83103817 cg03552151 -2.4012138 cg01538166 -4.8332151 cg03573179 -2.7746358 cg01557754 -4.61235 cg03588998 -0.4807417 cg01579218 -2.4021833 cg03604424 5.43833432 cg01597480 -2.8061322 cg03664992 31.8703656 cg01762785 0.2491522 cg03823084 -3.1743293 cg01791648 -5.1331644 cg03834467 -0.9860636 cg01835620 -4.7227796 cg03839949 -10.265735 cg01902704 0.30891281 cg03844971 -1.5663285 cg01971089 -5.7406146 cg03848890 -1.9066197 cg01988129 -0.2660887 cg03869874 4.16917666 cg02059055 6.14037906 cg03883502 10.5911797 cg02088403 -0.3011643 cg03887528 -0.3733555 cg02153490 -0.0921455 cg03950166 -1.558685 cg02161761 2.86628373 cg03982897 -0.4693096 cg02204442 -1.3815779 cg03986400 -1.2139964 cg02225085 2.25807193 cg04088674 -0.7946084 cg02232751 4.2091458 cg04129308 1.85528241 cg02254885 -14.127701 cg04154465 3.06782338 cg02256105 -0.020307 cg04157658 -1.7570169 cg02306162 -0.0499936 cg04229059 -6.4451535 cg02339392 -0.8155854 cg04267526 0.73174734 cg02361878 6.17892437 cg04270358 1.03213167 cg02462416 1.28635049 cg04338863 -3.3058175 cg02462487 -2.7352981 cg04407388 -1.6827815 cg02493740 -1.6079316 cg04445851 0.04240196 cg02501978 -1.5130389 cg04451175 -1.2048828 cg02722637 0.46587786 cg04508114 -0.4335248 cg02729030 -5.2153005 cg04512892 1.47577894 cg02763536 0.61200038 cg04531704 0.91127887 cg02767634 1.97373793 cg04673465 2.22081179 cg02867102 -10.428584 cg04742397 -4.6831153 cg02870946 -1.2303554 cg04753583 0.41593537 cg02942825 -2.7979306 cg04760708 0.70927954 cg02965178 2.52384557 cg04785213 0.70161645 cg03046819 -7.3084773 cg04786857 0.17092297 cg03092551 -0.1670602 cg04838627 -1.1294458 cg03155027 -0.0080934 cg04911050 4.68390918 cg03164928 -3.5957803 cg04998671 -30.960837 cg03167948 -0.1844817 cg05001334 1.03380228 cg03203114 -1.2488165 cg05003422 5.60587721 cg03227963 -0.0943263 cg05034363 2.92504981 cg03277049 5.124818 cg05059607 2.60469202 cg03283486 0.20334065 cg05070268 -1.2215682 term estimate term estimate cg05087948 1.1991173 cg06933824 3.45347204 cg05090759 -1.397405 cg06980387 5.82627257 cg05172940 2.42024622 cg06984176 0.14159407 cg05238695 -3.3973555 cg07155455 1.56473154 cg05260372 -1.7723352 cg07155684 2.89603282 cg05265042 -0.2320468 cg07186576 -0.3358436 cg05280698 4.0511649 cg07286682 0.68887153 cg05310309 1.55600268 cg07360805 -0.773523 cg05360774 1.25613217 cg07390013 0.81041383 cg05376617 0.55323002 cg07495704 -7.6333551 cg05395210 -4.1241958 cg07495811 2.41235283 cg05455729 -0.3270445 cg07560510 2.49582803 cg05463027 -14.37923 cg07657357 3.3756577 cg05470939 2.07754589 cg07671586 1.07640333 cg05561193 -1.3461417 cg07725123 1.05559048 cg05726118 2.53588419 cg07736657 -0.0639274 cg05861879 -4.0260888 cg07809027 -3.4112001 cg05874888 -0.8083869 cg07833467 -0.8148068 cg05900234 4.74763611 cg07850154 -24.234776 cg05922911 0.43707176 cg07910813 -1.0729767 cg05966235 0.13642299 cg07984980 -1.1511983 cg05980111 -1.7713828 cg08017858 0.82022065 cg05991454 7.601516 cg08025960 0.29773201 cg06007201 -12.449463 cg08046569 1.11084516 cg06024411 1.52945724 cg08081725 -0.9547729 cg06089468 3.46444719 cg08108311 8.88841119 cg06156376 4.91061933 cg08122369 -10.534933 cg06179486 -1.1384817 cg08129490 1.09065667 cg06275642 0.8324924 cg08166232 -0.428682 cg06449934 -0.175352 cg08170837 2.04109329 cg06470822 0.70317313 cg08173606 1.06622431 cg06493612 0.59124635 cg08190615 2.1362206 cg06574296 -0.0119487 cg08274097 2.64302111 cg06594770 4.51135658 cg08301612 -2.729191 cg06639733 -4.3949057 cg08317738 0.26269708 cg06658468 -2.1231767 cg08332662 -0.9323952 cg06670463 -4.5280541 cg08402963 0.65459904 cg06672696 10.5030019 cg08415508 -0.6785189 cg06675483 -0.9069373 cg08462924 2.21313438 cg06713116 -0.135586 cg08529529 -8.9566255 cg06734510 -8.9034084 cg08627089 0.12873143 cg06739520 5.13696838 cg08637514 2.15114206 cg06799422 5.01718809 cg08671671 1.33511191 cg06851000 0.93928715 cg08688335 -3.7960263 cg06882058 3.6706885 cg08733522 -0.2023862 cg06885782 -2.5846568 cg08762484 4.64062708 cg06916725 2.37227624 cg08797606 14.2895947 term estimate term estimate cg08826281 -2.6358023 cgl0958002 -0.2601732 cg08841511 2.65823287 cgl0960709 3.34339293 cg08863440 -1.9901855 cgl0975001 5.13128448 cg08916461 -0.8140773 cgl0999479 -0.0700323 cg08931376 -1.0839649 cgll053663 0.51629203 cg08965235 -1.7659968 cglll80122 0.38160951 cg09012544 11.0253743 cgll229399 5.00155235 cg09063262 -0.2829372 cgll244402 2.8480487 cg09164168 0.45031532 cgll326793 -4.2479654 cg09185587 -0.4486915 cgll369071 -0.4989324 cg09278098 -0.8205643 cgll524642 -0.3645636 cg09279566 -1.0027833 cgll545887 -0.3978629 cg09361966 -0.9050816 cgll573608 -0.5950011 cg09415366 1.02554887 cgll792186 -2.8525112 cg09450197 1.77830844 cgll835347 -6.0889257 cg09550397 -5.8053873 cgll846333 -6.070178 cg09573389 1.64146009 cgll946583 -1.2843143 cg09607276 0.94574236 cgll954355 -1.3845087 cg09662798 -2.7371179 cgll960655 -0.0902339 cg09896106 0.84267082 cgl2003463 -4.6515317 cg09906309 1.65079676 cgl2007048 -3.949833 cg09937438 -2.2196985 cgl2023170 -2.7268671 cg09974041 6.62067597 cgl2027899 1.58172957 cgl0046620 4.54479927 cgl2042659 -1.3789434 cgl0078511 -0.0416903 cgl2148898 -0.384638 cgl0110474 0.80717511 cgl2172441 2.25939625 cgl0243676 -1.6251679 cgl2179288 9.03260201 cgl0245988 1.55349551 cgl2211856 5.53003549 cgl0253371 3.96217418 cgl2212060 -1.8165233 cgl0406027 -3.2117192 cgl2226009 -0.2716606 cgl0421002 -1.3635941 cgl2257692 1.93368443 cgl0489614 -0.2639964 cgl2283398 1.16841617 cgl0515671 -6.0152328 cgl2316010 -2.4446777 cgl0529555 -3.6603521 cgl2387865 0.08803323 cgl0547057 -1.1200554 cgl2414301 1.42398185 cgl0557683 5.05242507 cgl2419195 9.72263643 cgl0577534 -5.5688057 cgl2419685 6.52999266 cgl0616300 3.87926028 cgl2419863 -1.7071248 cgl0619644 -0.8571093 cgl2614395 1.49723017 cgl0693071 0.47751499 cgl2666263 5.00932125 cgl0695490 -3.8320014 cgl2788037 4.95628137 cgl0715265 0.21336365 cgl2833018 1.64883965 cgl0750934 -1.9992943 cgl2908607 -0.4231708 cgl0755878 -0.0094494 cgl2978308 -1.6373962 cgl0809491 4.26757702 cgl3001893 3.04127886 cgl0923036 0.77527103 cgl3098855 0.4802255 cgl0951117 -0.8091901 cgl3202122 5.57624045 term estimate term estimate cgl3224583 4.62226563 cgl5964523 0.6993149 cgl3258563 -1.1869121 cgl6004055 4.4570457 cgl3444538 0.06770674 cgl6008966 -15.807545 cgl3483882 -1.7886747 cgl6080876 2.4362956 cgl3485809 12.169601 cgl6098332 0.15184917 cgl3511324 0.46275927 cgl6193278 -15.411232 cgl3561879 3.62771253 cgl6195091 2.59835513 cgl3569146 4.98657616 cgl6209444 -11.317501 cgl3665684 -3.7799892 cgl6248756 2.36742112 cgl3690424 23.2094324 cgl6312002 4.63900786 cgl3721134 3.62308515 cgl6321524 -4.5736367 cgl3798745 -0.0001335 cgl6427513 -2.0005936 cgl3813086 -0.1890568 cgl6511841 5.28105424 cgl3817265 -4.0834617 cgl6562257 2.85337053 cgl3826452 2.5553349 cgl6591681 -3.9963812 cgl3956645 3.95818583 cgl6633951 0.33813012 cgl3983063 -0.0171946 cgl6636110 -0.7183975 cgl4018471 -0.7749422 cgl6701167 0.23260948 cgl4067761 0.19897087 cgl6762979 -10.463298 cgl4095101 -1.4001818 cgl6810279 -4.4907529 cgl4241323 0.94905355 cgl6886581 0.30612081 cgl4593290 3.14809644 cgl6888547 -11.387985 cgl4611152 1.32505591 cgl6983588 -2.1515222 cgl4634687 4.47111967 cgl7092956 1.45618529 cgl4672293 13.5739468 cgl7263013 0.18858046 cgl4765414 4.14891241 cgl7272642 -2.4585006 cgl4848077 -0.767323 cgl7274064 -1.3867088 cgl4989252 -0.2316609 cgl7298973 1.94992 cgl5031579 2.70728813 cgl7304222 1.90388058 cgl5038286 -1.0398195 cgl7319774 0.99698296 cgl5046909 0.200946 cgl7344932 -5.5540327 cgl5086884 5.14813184 cgl7373751 -1.6554786 cgl5156071 4.58189083 cgl7390562 -8.6232068 cgl5205507 0.23859528 cgl7436666 2.18928834 cgl5213491 -0.9306521 cgl7459635 3.68913438 cgl5241130 -0.0182477 cgl7494199 -0.1715263 cgl5270892 8.25017133 cgl7514226 -5.6434186 cgl5299997 -5.1127891 cgl7516572 -0.4198843 cgl5383520 0.29117217 cgl7526103 2.03529389 cgl5443907 -0.969291 cgl7545662 -0.5688546 cgl5481172 0.61712916 cgl7576375 -9.294081 cgl5596301 0.36470394 cgl7646721 2.84486456 cgl5605172 1.6791391 cgl7658733 -4.7640179 cgl5622917 3.05433735 cgl7664577 -0.5925515 cgl5751090 1.97192522 cgl7681698 1.2490976 cgl5787227 -2.7295258 cgl7745234 -1.8125513 cgl5863539 2.20910781 cgl7848389 2.38731434 term estimate term estimate cgl7956485 -1.7032024 cg20816447 -2.6993394 cgl7968880 2.24821429 cg20856545 -0.4101868 cgl8050997 4.41984339 cg20861237 -0.279039 cgl8070470 2.53529387 cg20957370 3.95246859 cgl8137414 -0.7303666 cg21004924 0.70819353 cgl8180155 0.81790443 cg21121119 -1.1258615 cgl8196295 -0.5853404 cg21154793 -1.3037331 cgl8320379 -0.2215413 cg21160290 0.38957443 cgl8327056 9.46785717 cg21236593 -0.0652929 cgl8365211 -0.0793333 cg21249152 -1.5108204 cgl8468088 27.4876419 cg21293242 8.88572867 cgl8538662 2.79988528 cg21329085 3.77841251 cgl8644286 -0.9799952 cg21492308 1.9473482 cgl8735810 0.68244159 cg21527708 -0.0998074 cgl8775149 8.32237881 cg21571060 -3.4786234 cgl8797590 -5.8194081 cg21635854 5.12401889 cgl8958126 3.02207694 cg21642251 -2.4883914 cgl9039841 -0.8405036 cg21697134 -1.7557309 cgl9043574 3.33108161 cg21737698 2.0027006 cgl 9065831 0.27184155 cg21796167 -0.3492162 cgl9120897 -3.6694208 cg21904251 3.55752964 cgl9247841 14.723453 cg22013564 -7.5597036 cgl9261426 -6.075937 cg22040809 8.66741839 cgl9285688 4.08541936 cg22189725 7.14460451 cgl9399220 0.93454395 cg22202381 2.51265264 cgl9475108 0.06192032 cg22215631 3.379229 cgl9511338 -0.5194758 cg22225219 1.13297438 cgl9570154 0.41303613 cg22271663 0.87765391 cgl9692192 -2.571877 cg22277154 -4.8170056 cgl9812283 5.98507316 cg22284745 -0.6854184 cgl9935065 3.82916957 cg22442168 0.44928633 cgl9953038 3.98511867 cg22652782 -1.6209717 cgl9955500 -13.876876 cg22681495 2.3134632 cg20059012 -2.1235345 cg22697325 5.05111063 cg20141652 1.30401269 cg22698998 -0.9411975 cg20147046 4.1143005 cg22737282 0.86816672 cg20235117 -2.0613692 cg22761482 0.87556058 cg20245568 0.68005345 cg22807700 2.15317892 cg20320656 0.38267004 cg22872478 2.50465484 cg20326410 11.201295 cg22887526 -0.2401105 cg20368283 5.35696824 cg22889918 1.08548241 cg20494635 -3.9640405 cg22942200 -2.9175601 cg20532887 0.60259149 cg23065100 -0.6004525 cg20666917 2.18261687 cg23067299 -2.1841151 cg20704028 2.79610397 cg23112821 -3.6469509 cg20711218 3.41249842 cg23124451 -10.036495 cg20780880 1.00148751 cg23210521 1.44651388 term estimate term estimate cg23260993 3.89894021 cg25830305 0.47634254 cg23266598 -6.8488037 cg25893857 -0.4712614 cg23280730 -3.4125631 cg25932066 -0.4194662 cg23282585 -1.7229973 cg25945090 2.04267414 cg23285059 0.7204405 cg25956966 -1.7359694 cg23361092 10.3822605 cg25961618 0.9160339 cg23542533 1.026108 cg25979108 -0.7123194 cg23600866 2.0216572 cg26025543 -5.5626749 cg23626546 -8.707351 cg26070099 1.69812005 cg23634477 1.27451964 cg26084258 -9.3631758 cg23690166 -0.5875285 cg26168651 0.82825267 cg23698023 -12.985634 cg26235243 -1.6907373 cg23736055 -1.357795 cg26364871 -1.6094513 cg23788418 1.27724752 cg26365925 0.27623319 cg24010402 -0.2930937 cg26467949 -2.4875227 cg24158141 1.04632039 cg26635214 -6.3784403 cg24171453 2.75547998 cg26636010 -1.3822866 cg24479590 5.84523578 cg26780581 6.21876127 cg24670151 0.32400661 cg26795848 -0.3618613 cg24690437 4.35128473 cg26808167 0.75740509 cg24710309 0.30843471 cg26863750 -1.6048529 cg24741744 2.91739227 cg26888530 32.19979 cg24760922 -8.080979 cg26935333 -1.1980731 cg24768116 0.42211861 cg26936171 5.42365829 cg24784350 3.84589631 cg27021512 -9.0782511 cg24870774 8.22822408 cg27045062 -7.4391165 cg24891133 4.96581286 cg27051315 -6.2630039 cg24921858 -1.2505857 cg27096232 -1.1273398 cg24929896 -0.2920107 cg27175491 -0.9469938 cg24934400 4.43299139 cg27300045 -4.5334548 cg24949488 0.25582433 cg27321750 4.60763475 cg24952754 -1.9056489 cg27346545 -5.6040311 cg24977886 1.26836239 cg27355006 -3.4021639 cg24987259 -6.9437478 cg27379915 -0.8737653 cg25151919 0.95746929 cg27391693 -3.0055387 cg25152404 -0.2924506 cg27436995 2.16001599 cg25326896 25.0987055 cg27489373 1.18664998 cg25339052 -3.6695421 cg27516159 -2.4480191 cg25365379 -0.5510868 cg27529647 -2.6392064 cg25399541 -1.341181 cg27567593 -5.8765869 cg25473981 -10.150787 cg27587195 7.68513477 cg25519723 1.11150865 cg27631597 -0.435063 cg25645064 7.2631839 cg27646965 -9.0142653 cg25667997 -1.3546743 ch,17.1184801R 0.12060794 cg25732028 -0.8587237 ch.2.75889792R -1.5595759 cg25770948 0.9165665 ch.4.73355803R 1.41903713 cg25809722 -10.614293 ch.8.353716R 3.65921441 Table B - Dam Age CpG Sites and Estimates term estimate term estimate (Intercept) 543.431589 cg01082242 -18.039429 cg00003994 0.11111111 cg01091514 0.49710385 cg00023464 0.14209005 cg01103827 -1.0552951 cg00049440 0.8157786 cg01109734 -10.38002 cg00052482 -4.8587539 cg01136167 -0.3005004 cg00073543 0.47707559 cg01146808 -3.7485225 cg00084338 0.50020653 cg01161889 0.99876084 cg00115654 0.39611731 cg01168235 0.25733168 cg00117599 0.98554316 cg01205087 0.85419248 cg00192773 0.2023957 cg01220680 0.16150351 cg00228017 0.35894259 cg01221209 0.57827344 cg00296038 0.05865345 cg01302136 0.98595622 cg00300637 -1.7170812 cg01321673 0.27591904 cg00310410 0.60140438 cg01329511 0.85832301 cg00330279 -2.1801709 cg01439105 0.06443618 cg00332802 0.95662949 cg01449677 -2.9864245 cg00346985 0.4613796 cg01454752 0.11978521 cg00423487 0.87030153 cg01461718 0.04750103 cg00462168 0.16728625 cg01486146 0.41470467 cg00488692 0.06815366 cg01518465 0.44816192 cg00512563 0.15401732 cg01524149 0.22883106 cg00523379 -3.0033172 cg01557547 0.22222222 cg00534318 0.02684841 cg01557754 -14.19144 cg00554993 -7.1082145 cg01577414 0.40685667 cg00563845 -5.6154139 cg01597480 -13.20968 cg00603274 0.0425444 cg01603290 0.03148933 cg00612299 0.48533664 cg01615339 0.08178439 cg00614360 0.15448162 cg01619129 0.94671623 cg00645579 0.45600991 cg01659184 0.92234614 cg00655552 -4.90756 cg01682285 0.44609665 cg00697033 0.88599752 cg01687862 0.02395704 cg00717825 -2.1476926 cg01691194 -0.7867722 cg00720845 0.61214374 cg01711322 0.370095 cg00757033 0.44314717 cg01770362 0.16563404 cg00773060 0.34820322 cg01791648 -11.290423 cg00788025 -0.1715341 cg01831904 -0.3972666 cg00800780 0.22015696 cg01835620 -10.874175 cg00864474 0.15943825 cg01896085 -7.7666605 cg00877212 -3.0675184 cg01900832 -2.7591091 cg00894378 -0.7248034 cg01902704 0.98760843 cg00911370 0.42585708 cg01905210 0.07476249 cg00998451 0.34737712 cg01927000 -2.0163627 cg01005582 -2.6003224 cg01964856 0.32869366 cg01023759 0.32995896 cg02012043 0.16551544 cg01032119 0.35233375 cg02021288 0.40726972 cg01077274 0.54770756 cg02042310 0.23089632 term estimate term estimate cg02058357 0.98802148 cg03092551 -8.0041195 cg02059055 0.43659645 cg03162143 -6.4974051 cg02088403 -8.860954 cg03167948 -9.4438113 cg02171545 -0.6243651 cg03214087 -6.2794582 cg02192678 -0.8125605 cg03231447 -3.4719549 cg02208820 0.9748038 cg03283486 0.91449814 cg02218884 0.38289963 cg03286774 -1.0264384 cg02230495 -7.3797251 cg03303325 0.35398596 cg02232751 0.55679471 cg03336167 0.42213961 cg02244288 0.57954852 cg03338903 0.88806278 cg02247160 0.86864932 cg03360992 0.28831062 cg02254551 0.78314746 cg03508235 0.92523751 cg02254885 -31.562052 cg03519011 0.60305659 cg02256455 0.33126807 cg03520471 -13.57242 cg02283691 -3.6355907 cg03555424 -0.4071458 cg02339392 -13.600875 cg03565475 0.99834779 cg02414626 -9.5996382 cg03593550 0.22180917 cg02462416 0.78149525 cg03598731 0.57992565 cg02462487 -10.990453 cg03622371 -1.1468497 cg02464608 -0.8675449 cg03694580 0.89880215 cg02466947 0.00206526 cg03719092 0.51631557 cg02492920 0.08508881 cg03723356 0.90747625 cg02578470 0.24246179 cg03732007 0.59396943 cg02593958 0.49442379 cg03734594 0.33415944 cg02595575 0.10408922 cg03777083 0.88310615 cg02598071 0.41140025 cg03778594 -1.7054577 cg02610222 0.01486989 cg03780701 0.53531599 cg02693210 0.1928955 cg03783925 0.61090458 cg02697373 0.45022718 cg03805684 0.99958695 cg02704502 0.70879802 cg03817794 0.68938455 cg02762115 0.45146634 cg03843656 -0.6039877 cg02771117 0.47748864 cg03848483 0.23502685 cg02773041 0.33209418 cg03857047 -4.9085639 cg02794779 0.53655514 cg03869874 0.19537381 cg02822381 0.40437836 cg03882270 0.99421727 cg02825527 -1.5373353 cg03895593 0.54399009 cg02867102 -22.53154 cg03923640 0.22717885 cg02897366 0.24659232 cg03926598 -2.3249287 cg02952809 0.42709624 cg03950166 -14.369518 cg02955354 -3.9663651 cg03950599 0.10491532 cg02965712 0.4031392 cg03980370 0.68484097 cg02975922 -1.647494 cg03987653 -1.4263816 cg03000848 0.36735029 cg03990139 0.12616306 cg03021329 0.37092111 cg04000281 0.50185874 cg03036592 -5.8444584 cg04030848 0.42461793 cg03058664 -0.1974869 cg04038163 0.63403552 cg03071793 0.60057827 cg04042468 0.45642297 term estimate term estimate cg04091063 0.51094589 cg05248542 0.51548947 cg04120413 0.25361421 cg05265042 -8.9118833 cg04214075 0.07889302 cg05265359 0.42420487 cg04218812 0.21065675 cg05280698 0.08674102 cg04218880 0.21106981 cg05295671 0.12432879 cg04229059 -19.174025 cg05309877 -1.4987288 cg04231636 0.14869888 cg05324407 0.12224079 cg04307987 -1.8838152 cg05331334 -0.8602928 cg04322572 -0.3511388 cg05355167 0.86782321 cg04336659 0.48451053 cg05360774 0.28004957 cg04348250 0.12639405 cg05376617 0.08963238 cg04358463 0.30152829 cg05386977 0.63279637 cg04367197 0.11152416 cg05388545 0.93102024 cg04399631 0.31557208 cg05388821 -2.0595942 cg04407388 -6.3839574 cg05391998 0.41511772 cg04418999 0.99297811 cg05445326 -0.1731712 cg04445851 0.36761669 cg05463027 -29.740548 cg04474049 -1.42296 cg05483252 0.30524577 cg04505252 0.67410161 cg05521150 0.70218918 cg04528771 -5.803852 cg05577016 0.29904998 cg04571584 0.99669558 cg05593641 0.03841388 cg04655136 -0.2616534 cg05597836 0.46261875 cg04658021 0.82817018 cg05656900 0.75505989 cg04666465 0.35811648 cg05726118 0.05163156 cg04673465 0.41222635 cg05765580 0.13052458 cg04691795 -3.9202809 cg05767404 0.5633686 cg04694619 0.73667277 cg05770238 0.24204874 cg04717802 -0.327324 cg05774698 -2.3206316 cg04751549 0.54688145 cg05829145 0.17059067 cg04753583 0.50103263 cg05861879 -5.5436469 cg04781580 0.54812061 cg05863683 -0.2832814 cg04784327 0.7228418 cg05869537 0.06361008 cg04785284 -6.4082037 cg05896902 0.21520033 cg04788957 0.87360595 cg05980111 -5.5225434 cg04820362 -0.0635938 cg05991454 0.01404378 cg04845466 0.25857084 cg06024411 0.04295746 cg04872689 0.54729451 cg06073139 -1.2880606 cg04956585 -1.4331823 cg06120399 0.73853779 cg04995300 0.49153242 cg06145435 -0.6771877 cg05034363 0.01817431 cg06147863 0.37752995 cg05045517 -1.2198021 cg06156376 0.27344073 cg05048976 0.34613796 cg06182099 0.94630318 cg05055782 0.05080545 cg06204938 0.5377943 cg05102552 0.14787278 cg06217245 -2.2781315 cg05152300 0.85997522 cg06235390 0.51218505 cg05155047 0.98389095 cg06411551 0.09706733 cg05179172 0.52664188 cg06460691 -4.0771901 term estimate term estimate cg06473578 0.36637753 cg07255019 0.70962412 cg06490845 0.18215613 cg07286682 0.03593556 cg06504636 -1.3816152 cg07312601 -10.346276 cg06522772 0.72821148 cg07322898 0.11566731 cg06545268 0.11565469 cg07325246 0.27385378 cg06565975 -1.3893524 cg07379055 0.21354812 cg06594770 0.30028914 cg07379335 0.67988435 cg06619299 -1.6237109 cg07384080 0.39859562 cg06625004 0.14126394 cg07393255 0.09830648 cg06639733 -1.0880244 cg07401435 -7.3437923 cg06644488 0.3866171 cg07434944 0.35333896 cg06660332 0.28087567 cg07447773 0.26724494 cg06664254 0.38124742 cg07486199 -3.7348837 cg06682875 0.23973731 cg07495704 -23.247216 cg06697600 0.33002891 cg07537152 -0.4074002 cg06712651 0.9913259 cg07540084 0.53283767 cg06713116 -5.0008951 cg07547765 0.39363899 cg06723492 -1.8510846 cg07560510 0.15365551 cg06732989 -1.1033003 cg07571344 0.38496489 cg06734510 -3.6683441 cg07571928 0.27013631 cg06754224 0.67038414 cg07590529 0.20280876 cg06772202 0.57553486 cg07671586 0.32218092 cg06799422 0.3283767 cg07675337 0.39983478 cg06807593 0.88021479 cg07725123 0.40066088 cg06817264 0.05700124 cg07736657 -0.566537 cg06867482 0.54233788 cg07742235 0.67677982 cg06868100 0.45724907 cg07764386 -7.3464167 cg06871074 0.22800496 cg07779444 0.42337877 cg06872257 0.62866584 cg07800658 0.58653449 cg06872548 0.2432879 cg07803375 0.49896737 cg06882058 0.48409748 cg07833467 -0.4036801 cg06891458 0.21974391 cg07850154 -33.649864 cg06916725 0.00454358 cg07888957 0.19248245 cg06933824 0.25609252 cg07910813 -2.5746778 cg06980387 0.02808757 cg07917528 0.07369633 cg06995548 0.32796365 cg07925670 0.33952912 cg07000567 0.35952108 cg07938847 -2.6650278 cg07029024 0.874019 cg07961015 0.36596448 cg07097041 0.38950847 cg08030082 0.26559273 cg07104557 0.34151336 cg08034070 0.22635275 cg07163735 -7.3749949 cg08034171 0.16480793 cg07170253 -5.8797734 cg08046569 0.17554729 cg07200877 0.94052045 cg08081725 -16.501516 cg07216884 0.53366378 cg08096291 0.37505163 cg07235774 0.29574556 cg08129490 0.82321355 cg07235805 0.49194548 cg08203715 0.79471293 cg07240834 0.2094176 cg08208133 0.37918216 term estimate term estimate cg08220614 0.14327503 cg09119665 -4.0774313 cg08235413 -10.475345 cg09134314 0.93638992 cg08241514 0.4370095 cg09143673 0.377943 cg08248579 0.96860801 cg09151131 0.87980173 cg08270964 0.96158612 cg09153897 0.15489467 cg08274097 0.99256506 cg09164168 0.20693928 cg08277216 0.34944238 cg09206294 0.34696406 cg08301612 -10.993374 cg09234599 0.81990913 cg08317738 0.10097237 cg09294095 0.16026435 cg08349335 0.89384552 cg09323728 0.42668319 cg08373610 0.97315159 cg09328979 0.31722429 cg08402963 0.32094176 cg09363587 0.49938042 cg08428188 0.622057 cg09425279 0.16067741 cg08434127 -1.8528728 cg09428868 0.83767038 cg08443203 0.47459727 cg09521743 0.10574143 cg08467103 0.09845415 cg09547190 0.64229657 cg08526814 -24.671921 cg09578829 0.37339942 cg08529529 -27.936793 cg09628195 0.48161917 cg08530484 -1.030583 cg09645572 0.31882971 cg08551532 0.54702784 cg09728393 0.99008674 cg08583763 0.81825692 cg09754948 -15.802845 cg08593364 -17.422289 cg09773473 0.47418422 cg08614441 0.62701363 cg09832613 -1.4748338 cg08644365 -0.5085344 cg09837656 0.60718711 cg08649707 -5.2382722 cg09854088 0.4952499 cg08652441 0.37257332 cg09884146 0.93473771 cg08663634 0.43205287 cg09924848 0.7315159 cg08693738 0.02147873 cg09938213 0.29533251 cg08723357 0.98513011 cg09948192 -0.7247699 cg08733522 -7.2801308 cg09969462 0.36018174 cg08742502 0.98926064 cgl0046620 0.15737299 cg08749599 0.96076002 cgl0082647 0.19578686 cg08762484 0.67905824 cgl0110957 0.95704254 cg08764927 0.32259397 cgl0123952 -1.1268934 cg08797444 0.22470054 cgl0133725 0.47914085 cg08822136 0.20735233 cgl0147507 0.16811235 cg08824847 0.65551425 cgl0149296 0.35196017 cg08826281 -2.1134789 cgl0196532 0.07724081 cg08844900 -2.6351629 cgl0213353 0.42255266 cg08858130 0.08880628 cgl0314221 -3.862696 cg08878450 0.32266299 cgl0378538 -0.6561886 cg09012544 0.07063197 cgl0395519 0.54935977 cg09025327 0.47583643 cgl0406027 -9.6433505 cg09042411 0.20652623 cgl0426464 0.46840149 cg09053247 0.25145267 cgl0432859 0.59438249 cg09063262 -3.913876 cgl0460946 0.71623296 cg09110394 0.7306898 cgl0515671 -8.3004843 term estimate term estimate cgl0519437 0.03444374 cgll835347 -13.517496 cgl0529555 -6.4830674 cgll840849 -3.4640898 cgl0543574 0.10698059 cgll934819 0.37174721 cgl0599571 0.26352747 cgl2000995 0.23296159 cgl0612617 0.02933541 cgl2007048 -10.527821 cgl0616300 0.27426683 cgl2036877 0.52413672 cgl0662179 0.25898389 cgl2045999 0.18959108 cgl0693071 0.04832714 cgl2051614 0.31350682 cgl0695490 -4.952111 cgl2058385 -0.0018232 cgl0715265 0.17141677 cgl2119029 0.44444444 cgl0741153 0.50309789 cgl2122631 0.54316398 cgl0750934 -6.1174203 cgl2131894 -2.3862026 cgl0760299 0.23874432 cgl2133664 0.57290376 cgl0770076 0.49814126 cgl2188416 -0.7559321 cgl0780778 -1.5245162 cgl2205435 0.15184526 cgl0805511 0.118133 cgl2211856 0.38083437 cgl0809491 0.1842214 cgl2223258 0.13878563 cgl0836173 0.2180917 cgl2226009 -0.4107873 cgl0919204 0.4291615 cgl2283398 0.11482858 cgl0951117 -8.2730564 cgl2305200 0.05534903 cgl0960709 0.10656753 cgl2325455 0.25526642 cgll015497 -2.1695284 cgl2354986 -4.2564486 cgll029475 0.40603057 cgl2380854 0.73027675 cglll55735 0.14806705 cgl2459028 0.73275506 cglll73499 0.18746666 cgl2480416 0.22552664 cglll77223 0.7653862 cgl2491223 -12.299201 cgll218175 -1.0592031 cgl2580930 0.81206113 cgll290188 0.46798843 cgl2581592 0.24576621 cgll313708 0.20033044 cgl2594615 0.45105328 cgll358741 0.0842627 cgl2596182 0.71747212 cgll402700 0.16398183 cgl2606409 0.10037175 cgll421702 0.95993391 cgl2750151 0.90251962 cgll438134 -4.094798 cgl2753009 0.51135894 cgl 1449408 -1.1052523 cgl2788037 0.19041718 cgll471262 0.99752169 cgl2836280 0.30607187 cgll505841 0.37587774 cgl3000649 0.16191656 cgll517269 0.3738276 cgl3017983 0.72490706 cgll534593 0.29120198 cgl3058214 0.70838496 cgll540735 0.26807105 cgl3061373 -1.0841955 cgll562411 0.1771995 cgl3079123 0.41676993 cgll565355 0.08054523 cgl3127159 -1.9245147 cgll618577 0.13630731 cgl3127231 0.2779843 cgll663600 0.0243701 cgl3139020 -0.8313673 cgll756095 0.71045023 cgl3156931 0.51466336 cgll792186 -0.1382859 cgl3227806 0.15595552 cgll829633 0.91325898 cgl3283153 0.14085089 cgll835020 0.96034696 cgl3311096 0.25030979 term estimate term estimate cgl3323701 0.3527468 cgl4540297 -9.9491088 cgl3330671 0.1138824 cgl4584255 0.35609365 cgl3390332 -9.4188609 cgl4591667 0.13424205 cgl3393036 0.38042131 cgl4594111 0.95456423 cgl3399816 0.12102437 cgl4602471 0.52953325 cgl3417559 0.94134655 cgl4634687 0.92688971 cgl3435820 0.53903346 cgl4688451 -3.9222755 cgl3507964 0.40892193 cgl4692106 0.25691863 cgl3549152 0.58777365 cgl4701072 0.44286149 cgl3557773 -0.5718858 cgl4757738 0.25981 cgl3561879 0.01528294 cgl4768256 -0.9030108 cgl3569146 0.15324246 cgl4775751 -3.5926636 cgl3582001 0.4361834 cgl4781189 0.38455184 cgl3666174 -0.2186323 cgl4848077 -2.6341313 cgl3687915 0.9991739 cgl4903689 0.67781908 cgl3690424 0.1598513 cgl4905600 0.76290789 cgl3721134 0.19619992 cgl4919250 0.21685254 cgl3731523 0.13768293 cgl4939821 0.64849236 cgl3732582 0.53077241 cgl5031579 0.94589013 cgl3777609 -1.9780738 cgl5038286 -3.5037606 cgl3790268 0.49731516 cgl5063695 -14.618653 cgl3791379 -3.1212753 cgl5102179 0.51011979 cgl3792233 0.78769104 cgl5105011 -0.5760175 cgl3798745 0.3465923 cgl5232290 0.33587858 cgl3813086 -6.2266195 cgl5247329 -0.2113557 cgl3827984 0.5464684 cgl5258447 0.27302767 cgl3831329 0.40644362 cgl5262984 -1.7429628 cgl3868473 0.40850888 cgl5282632 0.629905 cgl3872065 0.85790995 cgl5299997 -13.147995 cgl3947929 0.04460967 cgl5373880 0.86947542 cgl3953458 0.09500207 cgl5384383 0.32011565 cgl4003022 -1.3985372 cgl5397472 0.33663775 cgl4053997 -0.1735621 cgl5409712 0.57703428 cgl4067761 0.8566708 cgl5431821 0.21643949 cgl4074486 0.39033457 cgl5445281 0.02886918 cgl4198472 0.23915737 cgl5543489 -0.0346824 cgl4228146 0.26765799 cgl5575356 0.88682363 cgl4242246 0.83188765 cgl5600051 -1.3099326 cgl4268226 0.81123503 cgl5604051 0.43907476 cgl4268632 0.24494011 cgl5611336 0.15076415 cgl4353201 -2.4924204 cgl5626112 0.5125981 cgl4363469 0.4118133 cgl5635368 0.88351921 cgl4368149 0.9834779 cgl5639684 0.14002478 cgl4371590 0.51920694 cgl5662902 0.1425031 cgl4388049 0.19206939 cgl5686393 0.06113176 cgl4442518 0.98182569 cgl5706250 0.19991739 cgl4524754 0.22511359 cgl5744005 0.35729038 term estimate term estimate cgl5751090 0.18793887 cgl6983588 -8.6006096 cgl5771128 -6.1777319 cgl7053538 -1.6703483 cgl5792487 0.01363073 cgl7054674 -8.427154 cgl5824291 0.30070219 cgl7088155 -0.4569773 cgl5840418 -0.0430472 cgl7105886 0.95415118 cgl5876676 0.39818257 cgl7109725 0.4978282 cgl5922176 0.92317224 cgl7173187 0.14952499 cgl5964523 0.24989674 cgl7179570 0.16935151 cgl5988970 -0.1499066 cgl7230535 0.41924824 cgl5996342 0.19950434 cgl7255214 0.43866171 cgl6010596 0.17802561 cgl7274064 -5.7760618 cgl6139227 0.35770343 cgl7279125 0.04667493 cgl6151795 0.36472532 cgl7279458 0.48864106 cgl6178415 0.64394878 cgl7310773 -5.8338111 cgl6195091 0.14622057 cgl7319774 0.04130525 cgl6213375 0.01776126 cgl7327990 0.29822387 cgl6218715 -9.0936119 cgl7334937 0.51301115 cgl6248756 0.63651384 cgl7430167 0.11028501 cgl6251130 -1.5017749 cgl7491146 0.36183395 cgl6296679 -0.5195628 cgl7494199 -3.8229739 cgl6308533 0.54275093 cgl7521665 0.98967369 cgl6326902 0.29987608 cgl7527798 0.47170591 cgl6335858 0.85501859 cgl7587327 0.22924411 cgl6368750 0.10739364 cgl7598574 0.39942173 cgl6375265 -3.3112668 cgl7667648 0.32465923 cgl6399833 0.73895085 cgl7672850 0.0691232 cgl6427513 -2.0526409 cgl7708016 0.44981413 cgl6448636 0.52209831 cgl7814814 0.2850062 cgl6457307 -0.6104153 cgl7852385 0.24617926 cgl6520312 0.03882693 cgl7870909 -8.312084 cgl6555466 0.87443205 cgl7877566 0.79223461 cgl6572224 0.76724479 cgl7910899 -0.0708525 cgl6596957 -0.9858203 cgl7951878 -3.9841042 cgl6643422 0.69516729 cgl7968037 0.4543577 cgl6653408 0.83973565 cgl7996830 -0.6127867 cgl6699385 0.44155308 cgl8034295 0.61296985 cgl6701167 0.22428748 cgl8059933 0.39487815 cgl6751098 -1.081061 cgl8064071 -0.4993625 cgl6824126 0.04956629 cgl8070470 0.26022305 cgl6845257 -0.4526925 cgl8071071 -0.5231766 cgl6861209 0.17472119 cgl8161890 0.39116068 cgl6886581 0.24080958 cgl8222590 0.78686493 cgl6888547 -7.6065654 cgl8245230 0.8409748 cgl6895261 0.28954977 cgl8257485 0.82651797 cgl6931969 0.06195787 cgl8297745 0.08467172 cgl6936289 0.9417596 cgl8320111 0.28748451 cgl6949914 0.9582817 cgl8329931 0.47666254 term estimate term estimate cgl8346576 0.97852127 cgl9462210 -1.5097708 cgl8365211 -5.0122056 cgl9506311 0.70590665 cgl8374181 0.24452705 cgl9514613 0.00123916 cgl8385671 -0.6639838 cgl9539664 -12.634464 cgl8419358 0.90541099 cgl9552640 -1.689361 cgl8449021 0.19867823 cgl9570154 0.26311442 cgl8468088 0.20363486 cgl9685479 0.05989261 cgl8477009 0.41003896 cgl9691410 -5.8936844 cgl8497052 0.19124329 cgl9697725 0.89591078 cgl8515886 -2.7680745 cgl9716643 0.46881454 cgl8538662 0.30194135 cgl9733534 0.49690211 cgl8552861 0.28624535 cgl9761014 0.97810822 cgl8581929 0.15778604 cgl9774627 0.20322181 cgl8611122 0.74266832 cgl9777853 0.43081371 cgl8625610 0.47046675 cgl9830657 0.1511772 cgl8634665 0.17265593 cgl9857407 0.07269723 cgl8638383 -2.7074454 cgl9900821 -1.6836541 cgl8668382 0.88104089 cgl9904425 0.15530772 cgl8674980 0.13093763 cgl9935065 0.47335812 cgl8696495 0.65097067 cgl9959917 0.4204874 cgl8735810 0.11648079 cgl9965810 0.07228418 cgl8737081 0.74060306 cg20014988 0.46964064 cgl8793806 0.31928955 cg20039814 0.43742255 cgl8797590 -8.2394651 cg20059012 -4.2804306 cgl8811731 0.58157786 cg20102877 0.1007848 cgl8833928 0.49979347 cg20110742 -9.9078836 cgl8894440 0.38372573 cg20120351 0.90458488 cgl8931760 0.39239983 cg20141652 0.10987195 cgl8940274 0.9574556 cg20155447 -0.5916918 cgl8958126 0.04874019 cg20172563 0.37959521 cgl9002763 -0.0251517 cg20203395 0.38166047 cgl9008597 0.75919042 cg20235117 -9.2745905 cgl9013753 0.17843866 cg20262330 0.29698472 cgl9021188 0.08937578 cg20271057 0.60019543 cgl9043574 0.06071871 cg20276402 -0.3678528 cgl9065831 0.30400661 cg20320656 0.92647666 cgl9066391 0.35852953 cg20321251 0.95786865 cgl9196221 0.2276792 cg20353653 0.92895498 cgl9197212 0.5212722 cg20356878 -0.0268829 cgl9205909 0.06567534 cg20433521 0.48120611 cgl9211382 0.755886 cg20454518 0.13919868 cgl9226100 0.98099959 cg20456258 0.23904239 cgl9248564 0.41057414 cg20468787 -1.3703523 cgl9308132 0.45229244 cg20482143 -2.59662 cgl9324714 0.39694341 cg20494635 -4.3340389 cgl9445684 0.61941408 cg20623702 0.88723668 cgl9452535 0.43494424 cg20631820 -0.1464032 term estimate term estimate cg20642413 0.74225527 cg21839331 0.07551636 cg20666917 0.0582404 cg21854228 -0.2803682 cg20701183 0.25939694 cg21923770 0.71750848 cg20708173 0.4109872 cg21926804 -0.3900075 cg20711218 0.0417183 cg21946667 0.34035523 cg20713174 0.1268071 cg21962450 0.1094589 cg20744163 0.15029271 cg22012583 0.87608426 cg20775810 -0.9542863 cg22022379 0.48203222 cg20780880 0.08921933 cg22025854 -1.8754215 cg20790367 0.43246592 cg22027946 -0.4842573 cg20797905 0.53159851 cg22079161 0.38248658 cg20802392 0.79677819 cg22103003 -2.9167224 cg20816447 -10.015135 cg22120714 0.00619579 cg20893838 0.28046262 cg22156842 -2.3737996 cg20908204 0.78479967 cg22189725 0.09624122 cg20991421 0.37835605 cg22202381 0.31598513 cg21004924 0.10326311 cg22239201 0.99586948 cg21052677 0.80503924 cg22242614 -3.9214824 cg21064451 -0.881377 cg22264409 0.39281289 cg21088119 -4.9299573 cg22283925 0.43411813 cg21112954 0.40933499 cg22348356 -10.320885 cg21136371 0.71086328 cg22425568 0.11400248 cg21144340 -0.4914244 cg22548220 0.23420074 cg21188242 0.82693102 cg22637538 0.01693515 cg21207665 0.39405204 cg22639561 -9.5440652 cg21248554 0.11317637 cg22652782 -18.139696 cg21251926 -1.5289413 cg22706610 0.50351095 cg21293242 0.2606361 cg22720431 -0.0843688 cg21320768 0.03634862 cg22733207 0.87897563 cg21333674 -2.0023163 cg22737282 0.00536968 cg21357291 0.41429162 cg22779878 -3.7872326 cg21415227 0.46509707 cg22826874 0.20487402 cg21436456 0.28170178 cg22860775 0.50846758 cg21479132 -2.6510691 cg22900229 0.36885584 cg21500300 0.74142916 cg22920538 -3.8021331 cg21500966 0.12143742 cg22935921 0.38744321 cg21571060 -7.4123425 cg22977317 0.79595209 cg21574853 -7.3866885 cg23030863 0.46179265 cg21644387 0.67327551 cg23043611 0.39570425 cg21672276 0.4535316 cg23065100 -0.5640633 cg21692450 0.01407262 cg23066280 -0.1532217 cg21697134 -1.790877 cg23080060 -1.922041 cg21759268 0.28211483 cg23124451 -13.492188 cg21782813 -4.3474898 cg23151014 0.51425031 cg21793437 0.94382487 cg23172400 0.41016109 cg21796167 -11.863581 cg23188704 0.51590252 cg21838488 0.11441553 cg23206745 -0.4351523 term estimate term estimate cg23207054 0.93143329 cg24617723 0.09582817 cg23210521 0.03180504 cg24650267 0.55762082 cg23260525 -0.5572306 cg24674269 0.45477076 cg23266598 -10.31931 cg24757926 0.52829409 cg23282585 -16.240948 cg24768116 0.10904585 cg23307798 0.08302354 cg24784350 0.08715407 cg23336797 0.55183808 cg24870774 0.00165221 cg23367683 0.15819909 cg24873872 0.35277505 cg23373153 -3.6167741 cg24874254 0.98017348 cg23464183 0.31268071 cg24913868 -10.011699 cg23489630 0.52168525 cg24920358 0.41635688 cg23581183 0.7984304 cg24924449 0.60760017 cg23581793 0.86327964 cg24935556 -2.8696772 cg23600866 0.26228831 cg24952754 -7.5999228 cg23613051 -0.131812 cg24983539 0.84262701 cg23618638 0.19496076 cg25036456 -2.4128653 cg23624713 -5.5455047 cg25119002 -0.2462162 cg23626546 -7.6915671 cg25122125 -4.4430393 cg23670519 0.72903759 cg25123427 0.79347377 cg23698023 -0.0562295 cg25127315 0.12928542 cg23736055 -4.5656728 cg25151919 0.49318463 cg23737061 0.87938868 cg25179758 0.91615035 cg23737927 -0.6394294 cg25188760 0.20948276 cg23777173 -0.8476144 cg25215230 -0.8463894 cg23799375 0.2771582 cg25326896 0.00660884 cg23830205 -3.296932 cg25351599 0.11358943 cg23833896 -1.8355147 cg25353171 0.32548534 cg23893460 0.23667906 cg25361506 0.26972325 cg23895495 0.62247005 cg25363789 -1.576559 cg24015175 0.99339116 cg25394782 0.12725713 cg24044052 -5.4747311 cg25409140 0.23213548 cg24087736 0.10473543 cg25481454 0.92028088 cg24109012 0.51886534 cg25486749 -1.803531 cg24112692 0.26187526 cg25509697 0.38826931 cg24164702 0.7645601 cg25561904 0.35687732 cg24240870 -0.5041117 cg25645064 0.04419661 cg24249248 -1.2904337 cg25660036 0.24039653 cg24253500 0.11832422 cg25671438 0.72779843 cg24311135 0.34985543 cg25697881 -0.4122874 cg24327262 0.53820735 cg25749107 0.00927456 cg24370881 0.48616274 cg25753817 0.88888889 cg24412006 0.71581991 cg25783326 0.24370095 cg24437311 -6.4799017 cg25846723 0.3692689 cg24453699 0.48905411 cg25860399 -0.9647031 cg24479590 0.00743494 cg25872744 0.40520446 cg24512005 0.61462206 cg25940248 0.55018587 cg24555670 0.26517968 cg25961618 0.77282115 term estimate term estimate cg25969992 -0.2306091 cg26693467 -4.0957978 cg26035892 0.71953738 cg26726141 -2.9017718 cg26070099 0.02560925 cg26734888 0.30978934 cg26082368 0.22098306 cg26780581 0.03469641 cg26097391 0.53448988 cg26781129 0.48244527 cg26157803 0.51714168 cg26808167 0.59231722 cg26160218 0.30701417 cg26848071 0.86245353 cg26175729 0.83932259 cg26850624 -0.2749982 cg26282236 0.42021274 cg26888530 0.20983065 cg26292895 -8.5276601 cg26951091 0.9487815 cg26307871 0.35481206 cg27021512 -12.094373 cg26317006 0.92854192 cg27022827 0.74019 cg26322872 -2.3475033 cg27039118 0.96819496 cg26325335 0.21957404 cg27045062 -6.4033385 cg26340700 0.77075589 cg27078652 -3.4844279 cg26365925 0.06030566 cg27227029 0.20528707 cg26397549 -6.2140885 cg27241134 -5.5578509 cg26403416 -0.9597117 cg27249858 0.88847584 cg26424649 0.2684841 cg27261733 0.36100785 cg26467949 -1.7064785 cg27262717 0.18876497 cg26468833 0.60388269 cg27300045 -2.7594005 cg26471982 0.63527468 cg27355653 -0.2932958 cg26493814 -0.8158753 cg27396824 0.25825019 cg26509915 -0.1443512 cg27434984 0.98719537 cg26514961 0.29491945 cg27532318 0.4548635 cg26562921 0.54068567 cg27574654 0.32300702 cg26635214 -15.180625 cg27629782 -1.1434961 cg26636010 -0.7634238 cg27637363 0.04584882 cg26684673 0.25237505 cg27645544 -0.210304

Table C - AdaptAge CpG Sites and Estimates term estimate term estimate (Intercept) -511.97428 cg00290758 -6.861E-05 cg00008671 -0.0137092 cg00295744 1.42084533 cg00017970 -0.0001084 cg00316485 -0.0016145 cg00048759 33.8029569 cg00335735 -5.526E-05 cg00050402 -0.0494625 cg00342891 -0.0037653 cg00089550 -0.0260158 cg00344422 -0.0001112 cg00099240 -0.000155 cg00346145 3.4042617 cg00108164 16.0467868 cg00388262 -0.0002657 cg00131893 1.96264444 cg00492070 -0.0003776 cg00158122 -0.000292 cg00505045 7.50388563 cg00223715 0.58293085 cg00513984 -9.961E-05 cg00229508 -7.096E-05 cg00539174 0.08913463 cg00277334 -0.001087 cg00544337 -0.0071478 term estimate term estimate cg00552753 -0.0005405 cg01595397 -8.643E-05 cg00561903 -0.0001166 cg01611548 -0.0007084 cg00577578 -0.0004865 cg01614478 0.73435215 cg00589581 -0.0112065 cg01641620 -0.0002175 cg00638945 -0.0159653 cg01647632 -0.0004737 cg00655982 -0.0052405 cg01676795 -0.0441272 cg00712841 -0.0017416 cg01678292 3.16798757 cg00715290 -5.667E-05 cg01686177 2.03776505 cg00750088 -0.000271 cg01707820 -0.0016673 cg00785170 -0.0011901 cg01768926 -0.0031455 cg00834400 -0.0003867 cg01783841 -9.586E-05 cg00851050 -0.0002258 cg01785233 -3.11E-05 cg00859280 -0.0003513 cg01787798 -0.000104 cg00870514 -0.0008508 cg01813672 -0.0002286 cg00877329 -0.0001899 cg01877606 -0.0002196 cg00910168 -4.542E-05 cg01899318 -0.0001192 cg00929523 -0.0001944 cg01943692 -0.0001723 cg00933182 -0.0002079 cg02010447 -1.964E-05 cg01019770 -0.0411924 cg02061130 -7.491E-05 cg01048752 -0.0071107 cg02061804 0.71145756 cg01055594 -0.0001758 cg02071712 -0.0001582 cg01065599 -0.0008653 cg02079584 -0.0002448 cg01080986 15.2784519 cg02131130 -0.0013704 cg01081263 -0.0001901 cg02133624 -3.241E-05 cg01103582 9.91943444 cg02145668 -0.0033739 cg01181940 -9.093E-05 cg02161761 1.74744658 cg01192291 -0.0015668 cg02186748 -0.0026756 cg01209296 -0.000947 cg02216481 -2.701E-05 cg01213022 6.54763092 cg02225085 10.3729591 cg01229452 -0.0246304 cg02264895 -0.0003963 cg01239922 -0.0053352 cg02320003 -0.0053547 cg01245393 -0.0002436 cg02361878 6.81994265 cg01262865 -7.843E-05 cg02393721 -0.0020756 cg01274524 4.98655078 cg02434059 4.4263351 cg01307174 -0.0001952 cg02435538 -0.0332023 cg01346077 -0.0003159 cg02450064 -0.0044686 cg01399860 -9.363E-05 cg02486497 -8.638E-05 cg01416891 -0.0003488 cg02491557 0.10150297 cg01421252 -0.000279 cg02493740 -0.0006179 cg01433677 -0.000813 cg02563156 -0.00027 cg01521220 -0.0186156 cg02569613 -0.0001494 cg01530283 -8.867E-05 cg02582963 -0.0007806 cg01534887 1.82962067 cg02653521 -9.223E-05 cg01538166 -0.0088581 cg02691360 -0.0018682 cg01544580 -0.0008566 cg02729030 -0.0002061 cg01563071 2.97887625 cg02761568 -0.009326 cg01579218 -0.0006783 cg02777885 -0.0001128 term estimate term estimate cg02780919 2.0397033 cg03748503 -0.0216813 cg02827075 -0.0019161 cg03755535 -0.0071813 cg02870946 -0.0047713 cg03787711 -9.557E-05 cg02942825 -0.0129901 cg03847705 -0.0001872 cg02954562 -5.749E-05 cg03858663 3.86639189 cg02965290 -0.0017316 cg03864121 5.0178558 cg02966722 -0.0029717 cg03871460 -0.017885 cg02967428 -0.0001142 cg03887528 -0.0007793 cg02995791 4.02307911 cg03948781 -0.0002827 cg03025337 -0.0149376 cg03982897 -0.0007042 cg03040622 -0.0015692 cg03995615 -0.0054975 cg03077331 -0.0001188 cg03999130 -0.0016614 cg03111404 11.2132307 cg04012082 -0.0001567 cg03140521 3.73788744 cg04013159 -0.0001653 cg03155027 -0.0123522 cg04035728 -3.333E-05 cg03165014 -0.0001446 cg04084236 -0.0026959 cg03177551 1.54347675 cg04087740 -0.0020312 cg03186975 -0.0034034 cg04115680 -0.0127436 cg03215416 1.80901321 cg04152629 -0.0014781 cg03270167 -0.0011071 cg04154465 2.68227346 cg03277049 12.8126008 cg04194821 -0.002043 cg03297163 -0.000168 cg04236639 -9.184E-05 cg03310376 2.87419955 cg04254769 -0.000109 cg03337277 -0.0002429 cg04259907 -0.0001812 cg03345116 -0.0018585 cg04270358 5.58002637 cg03405983 13.3725261 cg04292941 -0.0002322 cg03454541 -0.0030992 cg04295372 -5.812E-05 cg03466525 -5.835E-05 cg04297819 -4.461E-05 cg03493032 -0.0024895 cg04332818 11.2597771 cg03507218 0.49315792 cg04359828 -4.085E-05 cg03521737 -6.482E-05 cg04362886 -0.0002181 cg03525069 -7.545E-05 cg04365102 -0.0006964 cg03534847 8.48319645 cg04378886 -0.0046107 cg03537591 8.73685021 cg04452896 -0.0002107 cg03554174 -0.0008098 cg04531704 13.0054582 cg03573179 -0.0003573 cg04603184 -0.0280849 cg03574652 4.85122588 cg04613313 -0.0058607 cg03588998 -0.0007444 cg04654363 -0.0008042 cg03603381 -0.0134279 cg04677158 -0.000102 cg03639671 -0.0286827 cg04682845 -0.0001174 cg03641033 -0.0001127 cg04739880 -0.0580262 cg03669147 -0.0002479 cg04756296 -0.0009227 cg03678098 -5.27E-05 cg04760708 0.59400494 cg03686455 7.78826168 cg04764624 -0.0021552 cg03688058 -3.899E-05 cg04786857 5.90217557 cg03739378 -0.0006963 cg04872610 -0.0061115 cg03741653 -0.0001782 cg04889790 -0.001101 term estimate term estimate cg04897713 -0.0013779 cg05732876 -0.000465 cg04904276 -5.389E-05 cg05759421 7.48153639 cg04920452 -0.0002666 cg05787209 0.99914803 cg04928670 -0.0005847 cg05800368 -1.795E-05 cg05001334 6.52339754 cg05850205 -0.0008792 cg05003422 64.7347528 cg05874888 -0.0272041 cg05049335 -0.0001639 cg05890019 -0.0001798 cg05056497 -8.688E-05 cg05903289 0.30986479 cg05059108 -0.0001181 cg05922911 5.88774748 cg05059607 10.1022332 cg05929069 -0.0020583 cg05070268 -0.0004133 cg05951994 -0.0020771 cg05081614 -0.0009276 cg05957749 -0.0039017 cg05083128 -0.0004282 cg05977696 -0.0851238 cg05090127 -0.0019225 cg05985146 -0.0029178 cg05090759 -0.000525 cg06007201 -0.0002224 cg05106502 -0.0301172 cg06035815 -0.0009044 cg05131940 -0.0004133 cg06116806 -0.0011006 cg05132222 -0.0426715 cg06136160 -0.0005472 cg05147525 -0.0002729 cg06146665 -0.001896 cg05156137 -0.0002524 cg06197751 -0.0001116 cg05187965 -4.318E-05 cg06198975 3.41309839 cg05203213 -0.000615 cg06270615 -0.0005422 cg05208605 11.854737 cg06377473 -0.0004509 cg05290300 -5.809E-05 cg06407657 -4.395E-05 cg05310309 2.9479852 cg06412669 0.73105806 cg05323898 2.56600548 cg06418867 1.99003237 cg05339588 -0.003433 cg06496666 -0.0001511 cg05374271 -IE-04 cg06500246 -8.62E-05 cg05385434 -3.18E-05 cg06521852 -0.0001433 cg05395210 -0.0073297 cg06559878 -0.0002516 cg05399434 -0.0006185 cg06567829 5.61566519 cg05407338 -0.0145251 cg06573459 -0.0001583 cg05497175 -0.0002426 cg06599546 -0.0015005 cg05507697 -0.0121701 cg06606003 6.1442712 cg05517697 16.1079966 cg06629644 -0.0002048 cg05520031 -0.0001892 cg06636172 1.11185354 cg05523085 -0.0009532 cg06652313 3.30632514 cg05542681 6.69988285 cg06670463 -0.0153083 cg05551889 -0.0017211 cg06675483 -0.0247562 cg05561193 -0.0002567 cg06699216 0.09094129 cg05580441 7.8538039 cg06739520 64.8969906 cg05601974 -0.0004599 cg06806080 -0.0002464 cg05630016 -0.0002043 cg06864533 -0.0009026 cg05641529 5.81564598 cg06885782 -0.0002157 cg05673214 -0.0029182 cg06908352 -0.0001194 cg05709162 -0.0003372 cg06975196 -0.0011987 cg05724110 6.43544356 cg06984176 6.21451797 term estimate term estimate cg06994022 -4.224E-05 cg08131204 0.64877917 cg07003587 -0.0003245 cg08203210 -0.012207 cg07030727 -0.0001933 cg08248751 -2.204E-05 cg07053014 -0.0068286 cg08285446 -0.0004847 cg07068570 -0.0001402 cg08291907 -0.000357 cg07074042 -3.471E-05 cg08334034 -0.0001222 cg07077694 -0.0004451 cg08363339 -8.824E-05 cg07136905 -0.0006195 cg08438690 -0.0013078 cg07142010 -3.123E-05 cg08480461 -0.0038183 cg07186576 -0.0004449 cg08481354 21.123294 cg07196577 -0.000132 cg08516817 1.73841036 cg07203024 -0.0022773 cg08541155 -0.0002801 cg07244098 -9.885E-05 cg08577293 -0.0062754 cg07251046 2.71570422 cg08591668 -0.0002832 cg07326665 -0.0008775 cg08604223 -0.0048786 cg07330481 -0.0005948 cg08655800 -0.0009553 cg07356549 -0.0006608 cg08661007 6.49891285 cg07360805 -0.0021902 cg08665251 -7.234E-05 cg07364657 -5.356E-05 cg08690999 -0.0026184 cg07420274 -0.001397 cg08738300 2.0055399 cg07434260 -0.0022431 cg08759041 -0.0062415 cg07435237 19.1036172 cg08789022 4.33772383 cg07437373 -0.0023535 cg08832695 -0.0035916 cg07495811 2.3897303 cg08841257 -4.138E-05 cg07561710 -0.0018393 cg08897150 -5.471E-05 cg07571531 0.4676354 cg08915824 -6.608E-05 cg07597022 -0.0002009 cg08939521 3.32736941 cg07598052 -0.0001623 cg08965235 -0.005859 cg07599979 -6.439E-05 cg08980461 -0.0257429 cg07630301 -4.39E-05 cg09043511 -0.0011851 cg07637837 -0.0010589 cg09048038 -9.537E-05 cg07644368 -0.0001269 cg09144707 -0.0004531 cg07657357 -0.0047499 cg09173348 -0.0004748 cg07743805 -0.004286 cg09226596 4.5542474 cg07751331 -0.0001395 cg09230154 7.19588314 cg07834249 4.53266709 cg09236382 -0.000209 cg07890839 -0.0030152 cg09278098 -0.0015967 cg07907506 -8.545E-05 cg09293925 -0.0028738 cg07924503 -0.00039 cg09409484 -0.0056162 cg07932199 -0.00143 cg09438113 -9.695E-05 cg07951355 -0.0090853 cg09494188 11.956544 cg07977153 -0.0005822 cg09500200 -0.0050981 cg08017858 4.98898635 cg09550397 -0.003231 cg08022502 -0.0002409 cg09560549 -0.0004655 cg08027265 -0.0015001 cg09565806 -0.0014836 cg08084946 1.6632373 cg09582351 -8.713E-05 cg08110542 -8.5E-05 cg09584711 -0.0001069 term estimate term estimate cg09594075 -0.0017774 cgl0601821 -0.001271 cg09621438 -0.0034009 cgl0608948 -0.0001267 cg09650189 -0.0232144 cgl0619644 -0.0010473 cg09662798 -0.0127482 eg 10644544 -0.0001894 cg09700701 -0.0052909 cgl0660903 -7.895E-05 cg09729866 -0.0005413 cgl0708955 -0.0002336 cg09741713 -0.024343 cgl0722426 -0.000322 cg09754413 -0.0061506 cgl0738119 -0.0335186 cg09856367 -0.0012112 cgl0745272 -8.105E-05 cg09882118 -0.0093215 cgl0762466 -0.0001697 cg09904296 -0.0004129 cgl0852096 -5.116E-05 cg09921821 -0.0394451 cgl0853431 1.23990001 cg09933355 4.41143558 cgl0883038 -0.001899 cg09978077 -0.0594639 cgl0915716 -0.0007934 cg09996325 -0.0004074 cgl0975001 -0.0033364 cgl0018167 -0.000496 cgll004284 -0.0013133 cgl0084554 -0.0024193 cgll068337 -0.0413554 cgl0099638 -0.0040474 cgll069276 -0.0014785 cgl0101634 -0.0025154 cgll083280 8.53935096 cgl0103850 -0.0057709 cglll01109 1.35989504 cgl0116432 -0.0003199 cglll46034 -0.0005887 cgl0129391 -0.0208828 cglll46821 -0.0018657 cgl0208370 -0.0055845 cglll57584 4.89342914 cgl0230190 -0.0005722 cglll98589 -0.0022855 cgl0240139 1.15658674 cgll209249 -0.0002196 cgl0245988 10.4776609 cgll220565 -0.0001947 cgl0249997 -0.0008455 cgll229663 -0.000407 cgl0258419 -0.039464 cgll241549 1.76688515 cgl0259872 -0.0001507 cgll345976 -2.57E-05 cgl0324116 17.0587639 cgll386711 -3.389E-05 cgl0338112 -0.017777 cgll412468 -0.000133 cgl0339152 -0.0005657 cgll548083 -9.298E-05 cgl0362475 0.66527543 cgll565786 3.15495237 cgl0364115 -7.309E-05 cgll591636 3.02592675 cgl0374813 -0.0047356 cgll619602 -0.000898 cgl0395934 1.14958131 cgll682508 -8.372E-05 cgl0411590 -0.0146162 cgll682697 -0.0203438 cgl0441379 0.14473919 cgll754420 -0.0005461 cgl0461547 -0.0030334 cgll857646 -0.0027594 cgl0493259 -0.0101546 cgll881754 25.5170287 cgl0512089 -7.964E-05 cgll898958 1.18802409 cgl0516975 -0.0007033 cgll908751 -0.0907048 cgl0536276 -0.0001087 cgll988722 -0.0023231 cgl0547057 -0.007397 cgl2023170 -0.0008938 cgl0577534 -0.0011148 cgl2027899 12.3165515 cgl0585486 -0.0002869 cgl2084760 -0.0003542 cgl0589385 -0.0020643 cgl2183861 -8.361E-05 term estimate term estimate cgl2193345 -6.775E-05 cgl3456470 -0.0223684 cgl2198704 3.84488304 cgl3467814 -0.033056 cgl2234855 12.308298 cgl3483882 -0.0001351 cgl2236088 -0.0172554 cgl3501538 -0.0002546 cgl2307333 -3.134E-05 cgl3581582 3.98482316 cgl2334488 1.62290196 cgl3601739 -0.0004669 cgl2347757 -0.0167597 cgl3646005 -0.0007844 cgl2369353 -4.835E-05 cgl3656518 21.8594121 cgl2419195 1.25377797 cgl3665684 -0.00015 cgl2419863 -0.0001308 cgl3690564 -0.0014125 cgl2422683 -0.0001014 cgl3695646 -3.399E-05 cgl2435725 -0.0039013 cgl3711394 -0.0089063 cgl2447832 -4.735E-05 cgl3718185 -0.0001507 cgl2454167 -8.495E-05 cgl3779868 -0.0001508 cgl2513221 -0.0210837 cgl3793145 -0.0032572 cgl2564285 -0.0003132 cgl3809095 -0.0014863 cgl2565788 6.7615952 cgl3817265 -0.0003355 cgl2569592 -0.0155494 cgl3826452 17.2278413 cgl2582426 -0.0002311 cgl3845561 -9.585E-05 cgl2614395 21.2222197 cgl3849525 -9.496E-05 cgl2649238 -0.0025852 cgl3872005 -0.0155877 cgl2673499 5.96889479 cgl3881108 -0.0414679 cgl2676803 -0.0005171 cgl3882377 -0.0010323 cgl2701088 -0.0002908 cgl3924974 -0.0006399 cgl2776287 -0.0007655 cgl3945224 -0.0486116 cgl2781915 -0.0011424 cgl3952159 -0.0029825 cgl2797879 7.11475089 cgl3957413 -0.0001572 cgl2864912 -0.0002285 cgl3978542 8.66716859 cgl2903224 -7.715E-05 cgl3983063 -0.0001113 cgl2904135 -0.003768 cgl3993274 -0.0002069 cgl2916580 -0.0002504 eg 14044707 -0.0003826 cgl2968598 -7.593E-05 cgl4072027 -0.0023965 cgl3007701 -0.0002422 cgl4079463 1.35526782 cgl3026730 -0.001609 eg 14081465 -0.0012843 cgl3121938 -0.0001486 cgl4088282 -0.0002709 cgl3224583 3.65298153 cgl4124917 -0.0005915 cgl3247673 -0.0010167 cgl4150023 4.42920512 cgl3257412 -0.0048973 cgl4186846 -0.000685 cgl3261390 1.20235686 cgl4198450 -0.0001678 cgl3280882 2.81217335 cgl4202850 -0.026704 cgl3296238 -0.001971 cgl4233374 -0.0184175 cgl3303654 -0.0007252 cgl4242895 -0.0023335 cgl3331559 -0.0012942 cgl4248680 -0.0003074 cgl3363708 -0.0002274 eg 14311471 -0.0010345 cgl3375538 -0.0004199 cgl4329026 -0.0001717 cgl3432087 1.22122453 cgl4345882 -0.0015426 cgl3455704 -0.003375 cgl4397813 -0.0008411 term estimate term estimate cgl4409029 -0.0006048 cgl6224163 -3.449 E-05 cgl4434109 -0.0001689 cgl6238336 -0.0097638 cgl4544289 -0.0003594 cgl6321524 -0.0073329 cgl4570838 -9.671E-05 cgl6404106 -4.91E-05 cgl4611152 -0.0130718 cgl6408679 -0.0002478 cgl4615833 6.32625983 cgl6477774 -0.0002126 cgl4672994 -0.0009324 cgl6569937 0.03433592 cgl4682080 0.87907021 cgl6578883 9.57280333 cgl4697880 -0.0019134 cgl6624069 -0.0002074 cgl4752965 -0.0017098 cgl6675381 7.86339929 cgl4781394 -0.0001298 cgl6699148 -0.0006365 cgl4789828 -0.000403 cgl6759221 -0.0162095 cgl4907738 -0.0026835 cgl6795804 -1.736E-05 cgl4944923 -0.0001619 cgl6801102 -0.0683026 cgl4989252 -0.0002268 cgl6844661 -0.0004073 cgl5072306 2.90620111 cgl6987524 6.20581341 cgl5099418 -0.0013912 cgl7063929 7.02311545 cgl5123742 -0.0027246 cgl7076780 -0.0002294 cgl5149655 7.78576433 cgl7113968 -0.0198898 cgl5176213 -0.0007471 cgl7133183 1.98373155 cgl5246131 -0.0016665 cgl7207815 -0.0168735 cgl5298719 -3.112E-05 cgl7215446 -6.714E-05 cgl5308737 -0.0003995 cgl7217665 -0.0015454 cgl5337006 -0.0010463 cgl7359265 -4.767E-05 cgl5352315 -0.0002975 cgl7389077 -3.886E-05 cgl5461663 -0.0001275 cgl7402294 -0.00011 cgl5590153 -0.0002111 cgl7408380 0.0004082 cgl5594585 -0.0003945 cgl7414101 -8.218E-05 cgl5681737 -0.0079353 cgl7425144 -5.033E-05 cgl5763258 -0.000226 cgl7436666 1.02165451 cgl5770553 -8.556E-05 cgl7479898 -0.0008005 cgl5787146 -0.0002709 cgl7499941 -0.0342576 cgl5808924 -0.000348 cgl7506588 -0.0004731 cgl5841865 8.50528102 cgl7526103 23.8442804 cgl5856275 -0.0002432 cgl7545662 -0.0001344 cgl5884510 -0.0004901 cgl7607231 -0.0007984 cgl5890469 -9.461E-05 cgl7701035 3.42465059 cgl5907944 1.4974289 cgl7733447 -0.0019494 cgl6005271 2.2893694 cgl7737388 -0.0003246 cgl6008966 -0.0082816 cgl7745234 -6.215E-05 cgl6038868 -6.159E-05 cgl7768691 -9.069E-05 cgl6043345 -0.0001471 cgl7774395 1.48085437 cgl6103959 6.35898413 cgl7822955 -0.0001118 cgl6193278 -0.0118803 cgl7834136 -0.0005544 cgl6206504 -0.0015043 cgl7841545 -0.0026052 cgl6208357 6.06485793 cgl7880593 -0.0001726 cgl6209444 -0.005406 cgl7956485 -0.0024072 term estimate term estimate cgl8044113 2.652512 cgl9537511 -0.0253125 cgl8045859 1.18082014 cgl9630374 3.23212533 cgl8064468 -0.0944369 cgl9636519 -3.49E-05 cgl8087520 -0.0001246 cgl9704902 -0.0090227 cgl8116154 -6.638E-05 cgl9735514 -6.431E-05 cgl8180155 2.57701971 cgl9764489 0.62810655 cgl8183774 -0.0003991 cgl9766591 -0.0012732 cgl8247172 -0.0002149 cgl9776272 -0.0013346 cgl8267943 -0.0002648 cgl9782411 -0.0031379 cgl8282455 -0.0001063 cgl9801062 -0.0067036 cgl8282791 1.30763486 cgl9866195 -0.0202711 cgl8321051 -0.0003861 cg20116199 1.1357039 cgl8327056 43.5868352 cg20124188 -0.0010901 cgl8401778 -6.828E-05 cg20140333 -0.00017 cgl8427905 -0.0005622 cg20169823 -0.0001982 cgl8544564 -0.0009755 cg20185615 -0.0002476 cgl8559348 -0.0055978 cg20224218 -0.0003727 cgl8585107 -0.0001125 cg20250935 -0.0001614 cgl8618230 0.69050983 cg20359042 17.1283852 cgl8626035 -0.0001994 cg20364660 -4.241E-05 cgl8653195 -0.0003682 cg20375093 -0.000163 cgl8664965 -3.764E-05 cg20401567 -3.169E-05 cgl8675610 -5.964E-05 cg20471783 -0.0009638 cgl8742814 -0.0005056 cg20481287 -0.0001435 cgl8750833 -9.468E-05 cg20485144 -0.0008893 cgl8770216 -0.0012642 cg20630655 -0.0001081 cgl8815343 0.05105114 cg20651988 -0.0071973 cgl8825119 -0.000669 cg20651995 -0.0006424 cgl8954925 0.23873446 cg20668177 -0.0062858 cgl9003183 -0.0159285 cg20671920 -0.0224186 cgl9009417 9.54249755 cg20720686 -0.0493596 cgl9040266 -0.0009853 cg20811236 -0.0002026 cgl9065106 -0.0001724 cg20885815 -0.0125034 cgl9091930 -0.0098259 cg20891481 -0.0185041 cgl9115882 0.00208883 cg20918218 -0.0130935 cgl9120897 -0.00056 cg20918537 -0.0029454 cgl9162496 -0.0005801 cg20933634 -1.958E-05 cgl9220282 -7.129E-05 cg20957523 -0.0013369 cgl9261426 -0.0020876 cg20959174 -0.0140911 cgl9306368 -0.0002913 cg20981127 -0.0001326 cgl9308769 -0.0001318 cg21075829 -0.0139925 cgl9395684 -0.0001861 cg21093005 6.10425884 cgl9399220 8.61955651 cg21108553 15.3049011 cgl9413366 -0.0001143 cg21121119 -0.0052173 cgl9430473 -4.079E-05 cg21123203 -0.0038214 cgl9437258 -0.0072275 cg21139935 0.76268371 cgl9490598 10.2505587 cg21154793 -0.0006865 term estimate term estimate cg21161126 -0.0002536 cg22853588 -0.0003977 cg21220286 -0.0018809 cg22917652 -7.419E-05 cg21234538 -0.0133544 cg22927247 -0.0019981 cg21243944 -0.0001386 cg22968401 -0.0002374 cg21250433 -0.0009878 cg22987011 -4.724E-05 cg21274570 5.24734363 cg23003720 -0.0075371 cg21322248 -0.0002643 cg23012731 -0.0035634 cg21323720 -0.0010616 cg23075364 -0.0034308 cg21356710 -0.0002582 cg23090606 -0.0002289 cg21415385 -0.0009808 cg23112821 -0.0003224 cg21461308 1.64365821 cg23251687 -0.0779886 cg21467697 0.18652956 cg23273694 -0.0021948 cg21496419 -2.932E-05 cg23275972 -0.0003978 cg21602257 -0.0002784 cg23305365 -0.0002609 cg21635854 8.90734171 cg23307858 -5.12E-05 cg21639287 -0.000108 cg23322112 -0.0019918 cg21649013 -0.0027407 cg23333513 6.06323566 cg21672550 -0.007254 cg23361092 28.3906528 cg21684021 -0.0067514 cg23371476 -0.0014064 cg21689228 4.45839418 cg23435915 -8.616E-05 cg21709322 -0.006264 cg23445859 5.30497787 cg21737698 -0.0057604 cg23452458 -0.0022929 cg21748751 -0.0062904 cg23456221 -0.0006602 cg21761922 -7.216E-05 cg23463186 4.389869 cg21801165 -0.0017119 cg23469025 -0.0035193 cg21807240 -0.0009188 cg23550082 -0.0075443 cg21863334 -0.002387 cg23674344 -0.0005598 cg21886042 -0.0046166 cg23687466 -0.0004851 cg22013564 -0.0001408 cg23694533 -0.0008531 cg22017213 -0.0001402 cg23703060 -0.0001056 cg22021934 -0.0019246 cg23752007 -2.122E-05 cg22082347 -0.0004049 cg23765126 -0.00527 cg22158854 -4.854E-05 cg23817643 11.8252186 cg22160472 -0.0005102 cg23831735 -0.0140736 cg22217449 18.5309187 cg23939642 1.51259668 cg22271663 5.61557194 cg24012925 -3.868E-05 cg22277154 -0.0006998 cg24023487 -0.0001521 cg22326328 -0.0008419 cg24030449 -0.0004375 cg22360016 -0.0135373 cg24053748 8.21800989 cg22378853 -0.0014666 cg24057642 -0.0001166 cg22508957 -3.569E-05 cg24070213 -6.375E-05 cg22685966 5.37979139 cg24117274 -5.35E-05 cg22695707 0.03456224 cg24129115 -0.0029211 cg22708914 -0.0002942 cg24158141 4.56957546 cg22711218 -0.0018067 cg24159575 -0.0098242 cg22714777 -0.0035996 cg24217726 -0.0028829 cg22801992 -0.0001165 cg24251135 -0.0002752 term estimate term estimate cg24254377 -0.0001235 cg25822709 -0.0001013 cg24284497 -0.0004729 cg25848076 -0.0012687 cg24315876 -0.007627 cg25875213 -0.0447445 cg24322968 -0.0050598 cg25904183 -0.0001499 cg24387126 -4.558E-05 cg25912009 -0.0011492 cg24399540 0.45394521 cg25979108 -0.008859 cg24415799 -0.0004079 cg25982743 -0.0002782 cg24442609 -0.0007599 cg26004707 -0.0149953 cg24475171 -0.0001576 cg26012941 -0.0013209 cg24496614 -0.0238293 cg26034375 -8.728E-05 cg24514884 -0.0002707 cg26035489 -0.0005682 cg24609819 -0.0005618 cg26058502 -0.0002542 cg24636368 -0.0033239 cg26084258 -0.00814 cg24637417 -0.0003536 cg26109145 -0.0100455 cg24642820 -0.0001463 cg26124115 -0.0029991 cg24686551 -0.0632836 cg26127836 -0.0013067 cg24710309 7.03287791 cg26128121 -0.0011639 cg24727203 -0.0002095 cg26128464 1.3918397 cg24733384 -0.0002479 cg26134777 -0.0022724 cg24741744 -0.0013545 cg26137103 -9.524E-05 cg24749947 -0.0005326 cg26149658 -0.0001092 cg24761567 -0.0003084 cg26248173 -0.0018164 cg24765360 -0.0013858 cg26250086 -0.0001193 cg24787755 -0.0014955 cg26260789 -6.808E-05 cg24815934 -0.0001675 cg26261298 -0.0011119 cg24830730 -0.0047783 cg26276947 1.86575949 cg24856658 -0.0001345 cg26282505 -1.288E-05 cg24891133 9.78349506 cg26287345 -0.0005207 cg24928110 -0.0012956 cg26343958 -8.329E-05 cg24947451 -0.0001493 cg26425555 -1.488E-05 cg24950222 -0.0011538 cg26439710 -0.0074881 cg24987259 -0.0003173 cg26564280 -0.0011225 cg25064552 0.8056747 cg26572392 -0.0005166 cg25164649 -0.0522758 cg26578149 -0.0001731 cg25284762 -0.0039331 cg26644052 -0.0006888 cg25352397 0.64203147 cg26680047 4.52886414 cg25359664 -0.0003261 cg26684319 -2.705E-05 cg25428553 -0.0132 cg26692296 6.00575743 cg25557995 3.2690613 cg26692822 -0.0007135 cg25615068 -0.0239951 cg26749414 1.17194167 cg25618559 7.09091958 cg26767214 -0.0031972 cg25671484 -5.211E-05 cg26782013 -3.483E-05 cg25673241 13.5426927 cg26784012 -0.0001495 cg25741837 -0.0025648 cg26815396 -0.0019934 cg25744957 -0.0003404 cg26843498 -3.808E-05 cg25814293 -0.0095645 cg26863750 -0.0040026 cg25815229 -0.0075277 cg26875135 -8.449 E-05 term estimate term estimate cg26884773 -0.0006079 cg27222157 3.1391154 cg26901111 -0.0005769 cg27292417 1.70093743 cg26916966 -0.0860738 cg27304328 4.05536228 cg26940479 -0.0011618 cg27342333 -0.0302718 cg26951440 -8.971E-05 cg27346545 -0.0012903 cg26971710 -0.0070676 cg27355006 -0.0074673 cg27000590 -0.000167 cg27368025 -9.058E-05 cg27004870 -0.0030244 cg27379915 -0.0011042 cg27089226 -0.000235 cg27413008 -8.407E-05 cg27130359 -0.0012633 cg27587195 61.1972799 cg27134322 2.20380511 cg27598107 -0.0005066 cg27139956 -0.0001189 cg27598956 -0.0021883 cg27144223 -8.596E-05 cg27615366 -0.0007188 cg27184585 -0.003216 cg27631597 -0.0037344 cg27189341 -4.542E-05 cg27660099 2.94705724 cg27189533 -0.0006776 cg27661460 -6.912E-05 cg27208169 -0.0003438 ch.l.237398078F -0.0026718

EXAMPLES

The invention is further described in the following examples, which do not limit the scope of the invention described in the claims.

Methods

The following materials and methods were used in the Examples set forth herein.

Framingham Heart Study Cohort

The FHS cohort 1 is a large-scale longitudinal study started in 1948, initially investigating the common factors of characteristics that contribute to cardiovascular disease (CVD), firaminghamheartstudy.org/index.php. The study initially enrolled participants living in the town of Framingham, Massachusetts, who were free of overt symptoms of CVD, heart attack or stroke at enrollment. In 1971, the study started the FHS Offspring Cohort to enroll a second generation of the original participants’ adult children and their spouses (n= 5124) to conduct similar examinations. Participants from the FHS Offspring Cohort were eligible for our study if they attended both the eighth examination cycle and consented to having their molecular data used for the study. We used 2,544 participants from the group of Health/Medical/Biomedical (IRB, MDS) consent with available DNA methylation array data. The FHS data are available in dbGaP (accession number: phs000363.vl6.pl0 and phs000724.v2.p9).

Deaths among the FHS participants that occurred prior to January 1, 2013 were ascertained using multiple strategies, including routine contact with participants for health history updates, surveillance at the local hospital and in obituaries of the local newspaper, and queries to the National Death Index. Death certificates, hospital and nursing home records prior to death, and autopsy reports were requested. When cause of death was undeterminable, the next of kin were interviewed. The date and cause of death were reviewed by an endpoint panel of 3 investigators. Peripheral blood samples were collected at the 8th examination. Genomic DNA was extracted from buffy coat using the Gentra Puregene DNA extraction kit (Qiagen) and bisulfite converted using the EZ DNA Methylation kit (Zymo Research Corporation). DNA methylation quantification was conducted in two laboratory batches using the Illumina Infmium HumanMethylation450 array (Illumina). Methylation beta values were generated using the Bioconductor minfi package with Noob background correction.

Women’s Health Initiative

The WHI is a national study that enrolled postmenopausal women aged 50-79 years into the clinical trials (CT) or observational study (OS) cohorts between 1993 and 1998 ⁶ ’ ⁷ . We included 2107 WHI participants with available phenotype and DNA methylation array from “ Broad Agency Award 23” (WHI BA23). WHI BA23 focuses on identifying miRNA and genomic biomarkers of coronary heart disease (CHD), integrating the biomarkers into diagnostic and prognostic predictors of CHD and other related phenotypes. The death status was based on the variable DEATHALL (All Discovered Death) as listed in the form “All Discovered Death Outcome Detail (Form 124/120)”, generated on March 1, 2017. This variable does not censor deaths that occur after the participants’ last consent period. The original WHI study began in the early 1990s and concluded in 2005. Since 2005, the WHI has continued as Extension Studies (Extl), which are annual collections of health updates and outcomes in active participants. The second Extension Study (Ext2) enrolled 93,500 women in 2010 and follow-up of these women continues annually. Death was adjudicated for clinical trial (CT) and observational study (OS) participants through Extl. In Ext2, death is only adjudicated for the Medical Record Cohort (MRC). Non-MRC cause of death is determined by the initial cause of death form (form 120). DNA methylation quantification

In brief, bisulfite conversion using the Zymo EZ DNA Methylation Kit (Zymo Research, Orange, CA, USA) as well as subsequent hybridization of the HumanMethylation450k Bead Chip (Illumina, San Diego, CA), and scanning (iScan, Illumina) were performed according to the manufacturers protocols by applying standard settings. DNA methylation levels (P values) were determined by calculating the ratio of intensities between methylated (signal A) and un-methylated (signal B) sites. Specifically, the P value was calculated from the intensity of the methylated (M corresponding to signal A) and un-methylated (U corresponding to signal B) sites, as the ratio of fluorescent signals P = Max(M,0)/[Max(M,0)+Max(U,0)+100], Thus, P values range from 0 (completely un-methylated) to 1 (completely methylated).

DNA methylation quantitative trait loci (meQTLs) cis-meQTLs used in the study were obtained from the Genetics of DNA Methylation Consortium (GoDMC). DNA methylation levels were measured in whole blood samples from 36 cohorts, including 27,750 European subjects. In total, 420,509 CpG sites were analyzed to map the genetic influences on DNA methylation levels. The cis-acting meQTLs were defined as meQTLs within 2 MB window around the target CpG site. GoDMC summary statistics are available at mqtldb.godmc.org.uk.

GWAS data for DNA methylation and aging-related phenotypes

The twelve aging-related traits examined in the study include two lifespan- related traits (lifespan and extreme longevity) ^48,49 , three health-related traits (healthspan, frailty index, and self-rated health) ^50,5 four epigenetic age measurements (Horvath age, Hannum age, PhenoAge, and GrimAge) ⁵² , and three summary-level aging-related traits (Aging-GIPl, adjusted Aging-GIPl, and healthy aging) ^53,54 .

For the two traits related to lifespan, the parental lifespan GWAS included a total of 512,047 mothers and 500,193 fathers of European ancestry. For GWAS, the parental lifespan was equivalent to the lifespan of individuals, since the genetic effect on a parental phenotype is expected to be half of the individual’s phenotype itself ⁴⁸ . The extreme longevity GWAS included 11,262 subjects of European ancestry with a lifespan above the 90th percentile as the case group and 25,483 control subjects whose age at the last visit was below the 60th percentile age ⁴⁹ .

For the three health-related traits, healthspan was defined as the age of the first incidence of any major age-related disease, including dementia, congestive heart failure, diabetes, chronic obstructive pulmonary disease, stroke, cancer, myocardial infarction, as well as the incidence of death ⁵⁰ . The GWAS of healthspan included 300,447 subjects of European ancestry from the UK Biobank cohort, aged 37 to 73. The frailty index was calculated based on the cumulative number of health deficits during aging ⁵¹ . The frailty index GWAS included 164,610 UK Biobank participants aged 60-70 years and 10,616 Swedish TwinGene participants aged 41-87 years ⁵⁵ . Self-rated health GWAS was based on questionnaire responses on a scale of 0-5 in the UK Biobank cohort.

For the four epigenetic age measurements, the epigenetic age was based on various aging clock models, including Horvath age (353 CpG sites), Hannum age (71 CpG sites), Pheno Age (513 CpG sites), and Grim Age (1,030 CpG sites), are calculated in 34,710 participants of European ancestry ⁵² . All summary statistics of GWAS are publicly available.

For the three summary-level trait, the Aging-GIPl is the first genetic principle component of six human aging traits - healthspan, father and mother lifespan, exceptional longevity, frailty index and self-rated health, which captures both length of life and indices of mental and physical wellbeing ⁵³ . The Aging-GIPl -adj is the aging-GIPl adjusted for household income and socioeconomic deprivation. The Healthy Aging is the meta-analysis of healthspan, lifespan, and longevity ⁵⁴ .

Genetic correlation analysis

Genetic correlation between traits related to aging is calculated using the LD score regression (LDSC) ⁵⁶ . SNPs that were imperfectly imputed with INFO less than 0.9 or with a low minor allele frequency less than 5% were removed to reduce statistical noise. LDSC was performed using LDSC software vl.0.1 (github . com/bulik/ldsc) .

Epigenome-wide Mendelian Randomization analysis

In MR analysis, the definition of causal relationship is that associations of SNPs with CpG methylation are directionally consistent and proportional in magnitude to associations of SNPs with aging-related phenotypes. Genetic variants that are strongly associated with whole blood DNA methylation level (FDR < 0.05) were used for the MR analysis. Only meQTLs in the cis-acting regions were used to avoid pleiotropic effects. As the generalized MR method achieves a higher statistical power by including partially correlated instruments while accounting for the LD structure, we used LD clumping to only remove meQTLs with strong LD (r ² > 0.3), as suggested by Burgess et al. ³⁴ . Three MR methods were used for the main analyses: Wald ratio when only one meQTL was available, generalized inverse variance weighted (gIVW) when at least two meQTLs were available, and generalized MR- Egger regression (gEgger) when at least three meQTLs were available ³⁴ > ⁵⁷ > ⁵⁸ . The MR analyses were conducted using the MendelianRandomization R package and TwoSampleMR R package (github.com/MRCIELl/TwoSampleMR) ^59,6 °.

We only included cis-meQTLs (meQTLs located within 2 MB of target CpG sites) in our analysis to avoid pleiotropic effects, as they are more likely to affect DNA methylation via direct mechanisms. To remove additional pleiotropic effects, we used the results of gEgger, whose estimate is robust to directional pleiotropic effects if the significant intercept is detected by gEgger regression ( < 0.05).

CpG-phenotype pairs with Adjusted < 0.05 after Bonferroni correction were used to select causal CpG sites with the strongest MR evidence. All CpG-phenotype pairs with FDR < 0.05 were considered potential causal CpG sites and used in the downstream sensitivity analysis.

Sensitivity analyses

Horizontal pleiotropy. The horizontal pleiotropic effect in instrumental variants may cause biased causal effect estimation from the gIVW method. To detect unbalanced horizontal pleiotropy among genetic instruments, we used the intercept gEgger regression, which provides an estimate of the directional pleiotropic effect ⁶¹ . Note that by including a partially correlated instrument, the gEgger intercept also has more statistical power to detect the pleiotropic effect. CpG-phenotype pairs with gEgger intercept P value less than 0.05 were potentially affected by the pleiotropic effect, and the gIVW method may be biased. We, therefore, reported an estimate and P value from the gEgger method instead of the gIVW method for these MR signals, as the gEgger estimate is robust to the horizontal pleiotropic effect ⁶¹ .

Heterogeneity. To detect heterogeneity of the MR estimates in each meQTL, we performed the Cochran's Q test and the Rucker's Q test for the gIVW and gEgger results, respectively. Since heterogeneity does not necessarily affect causal effect estimation, we kept the MR signals heterogeneous while reporting potential heterogeneity in the result table.

Directionality test. To exclude MR signals caused by reverse causality (i.e., methylation changes caused by outcome phenotype), we applied MR Steiger test ⁶² , which is the method to test the directionality of causal effect estimated by MR. We then removed all MR signals with reverse directionality.

Colocalization analysis. Genetic colocalization is a Bayesian approach that estimates the probability (PP.H4) of overlapping genetic signals between molecular traits and outcome is due to both traits sharing a causal variant ⁶³ . It is an important method to control false positive results from MR and filter out the MR signals purely driven by LD or pleiotropy. All MR signals that passed the FDR threshold of 0.05 were then subjected to the colocalization analysis. We applied pairwise a conditional and colocalization (PWCoCo) analysis ⁶⁴ , which is a powerful genetic colocalization approach that is able to detect multiple independent genetic signals. We considered colocalization probability (PP.H4) > 70% as strong evidence of colocalization. Also, since aging-related GWAS are in general noisy while cis-meQTL usually have strong genetic signals, colocalization probability tends to be low, and the probability of only having a signal from meQTLs (PP.H1) tends to be high. To overcome bias due to imbalanced power between exposure and outcome traits, we considered a conditional colocalization probability (conditional PP.H4 = PP.H4 / PP.H3+H4) by assuming that the aging-related trait always has genetic signals in the region when a significant MR signal is detected. We then also reported CpG-phenotype pairs with conditional PP.H4 > 70% as a potentially colocalized signal.

Mediation analysis

We conducted multivariable MR (MVMR) to dissect significant CpG- phenotype causal effects (0T) into direct effects (0D) and indirect effects through transcript levels following the methodology outlined in Sadler et al., 2022 and using the smr-ivw software (github.com/masadler/smrivw) ⁶⁵ . Genetic effect sizes on CpGs (mQTLs) came from the GoDMC consortium (N=32,851) ⁴⁵ , and on transcript levels (eQTLs) from the eQTLGen consortium (N=$31,684) ²¹ , both derived from whole blood. Mediation analyses were assessed for CpG-Aging-GIPl and CpG-adjusted Aging-GIPl pairs.

Transcript mediators were selected to be in cis (< 500kb away from the CpG site) and causally associated to the CpG. This latter condition was satisfied when univariable MR effects from the CpG site on the transcript had an MR p-value < 0.01. Instrumental variants were required to be associated to either the CpG or included transcripts and as in the univariable MR analysis they were selected to be correlated at r ² < 0.3. The mediation proportion (MP) was estimated as 1- 0D/ 0T.

Causality-informed epigenetic clock model

Elastic net regression is a regularized linear model that solves the problem. where RSS( ) is the residual sum of squares, A is the regularization parameter, and /(/?) is the regularization term ⁶⁶ . The /(/?) term is the sum of the LI and L2 terms, which is defined as

The parameter is the elastic net mixing parameter, which controls the balance between the LI and L2 terms, wy is the penalty factor for each feature we introduced. In a regular epigenetic clock model, wy is defined to be 1 for all the features, which produces the model that is purely based on correlation ^67,68 . We introduced a causality-informed elastic net model, where we defined the feature-specific penalty factor as

PSf wy =

Sf = (max(cy) + min(cy) — cy) ^T

Here the c is the absolute value of the causality score for each feature, which is calculated from the causal effect size from MR weighted by colocalization probability. The T > 0 is a tuneable parameter that controls how much the causality score affects the feature-specific penalty factor. If T = 0, the whole model is reduced to a regular elastic net regression, where uy = 1 for all the features. When T becomes large, the model is more influenced by the causality score and tends to assign larger coefficients to the features with a higher causality score. To balance the precision and causality, we defined T as 0.3, which is the largest T value with MAE < 5 years in the validation set and maximized the association with mortality.

Using this method, we trained the model on whole-blood methylation data from 2,664 individuals ³⁵ . We built the causality-informed epigenetic clock model CausAge (See Table A). To separately measure adaptive and damaging DNA methylation changes during aging, we further separated the causal CpG sites into two groups based on causal effect size from MR and the direction of age-related changes. We then built Dam Age. _j a damaging clock (see Table B), and AdaptAge, a protective clock (see Table C).

Mortality association analysis

To evaluate our novel causal clocks for predicting all-cause mortality risk, we applied the clocks to a large-scale dataset comprising 4,651 individuals from (a) the Framingham Heart Study, FHS offspring cohort (n = 2,544 Caucasians, 54% females) ^48,49 and (b) Women’s Health Initiative cohort ^50,51 (WHI, n = 2107 postmenopausal women). Methylation levels were profiled in blood samples based on Illumina 450k arrays. In FHS, the mean (SD) chronological age at the time of the blood draw was

66.3 (8.9) years old. During follow-up, 330 individuals died. The mean (SD) followup time (used for assessing time-to-death due to all-cause mortality) was 7.8 (1.7) years. The WHI cohort is a national study that enrolled postmenopausal women aged 50-79 years. Our WHI data consists of three ethnic/racial groups: 47% European ancestry (Caucasians), 32% African Americans, and 20% Hispanic ancestry. All the three ethnic groups have marginally the same age distribution, with a mean (SD) of

65.4 (7.1) years old. The mean (SD) of follow-up time was 16.9 (4.6) years. During follow-up, 765 women died. To evaluate our clocks, we first defined age acceleration (AgeAccel) measure using the residuals resulting from regressing the DNAm variable on chronological age. As noted, this AgeAccel measure is independent of chronological age. Next, we applied Cox regression analysis for time-to-death (as a dependent variable) to assess the predictive ability of our causal clocks for all-cause mortality, using the AgeAccel measures. The analysis was adjusted for age at the blood draw and adjusted for gender and batch effect in FHS. We stratified the WHI cohort by ethnic/racial groups and combined a total of 4 results across FHS and WHI cohorts by fixed effect models weighted by inverse variance. The meta-analysis was performed in the R metafor function.

Example 1. Epigenome-wide Mendelian Randomization on aging-related phenotypes

MR is an established genetic approach for causal inference that utilizes natural genetic variants as instrument variables. Since the allocation of genetic variants is a random process and is determined during conception, the causal effects estimated using MR are not biased by environmental confounders. Therefore, it could be used as a tool for investigating causal relationships between the DNA methylation and aging- related phenotypes (FIG. la). In the context of MR, a CpG site can be defined as causal when associations of SNPs with CpG methylation are directionally consistent and proportional in magnitude to its associations with aging-related phenotypes.

To identify CpG sites causal to aging, we used 420,509 CpG sites with meQTLs available (GoDMC, whole blood samples from 36 cohorts, 27,750 European subjects) as exposures and selected twelve aging-related phenotypes as outcomes (FIG. la, Methods, Table 1), including two lifespan-related traits: lifespan and extreme longevity (defined as survival above 90th percentile) ⁵² ; three health-related traits: healthspan (age at the first incidence of any major age-related disease), frailty index (measurement of frailty based on the accumulation of a number of health deficits during the life course), and self-rated health (based on the questionnaire responses) ^53,54 ; four epigenetic age measurements (Horvath age, Hannum age, PhenoAge, and Grim Age) ⁵³ ; and three summary-level aging-related traits: Aging- GIP1 (the first genetic principle component of six human aging traits - healthspan, father’s and mother’s lifespan, exceptional longevity, frailty index and self-rated health), socioeconomic traits-adjusted Aging-GIPl, and healthy aging (multivariate genomic scan of healthspan, lifespan, and longevity) ⁵³ .

Aging-GIPl captures both the length of life and age-related health status ⁶⁹ , which can be considered as a genetic representation of healthy longevity. It also shows the strongest genetic correlation with all other traits related to lifespan ⁵³ . Therefore, we further used Aging-GIPl as the primary aging -related trait to investigate CpG sites causal to the aging process. A genetic correlation analysis showed that all eight lifespan- and health-related traits are genetically correlated and clustered with each other, while the four epigenetic age measurements clustered with each other. GrimAge and PhenoAge showed significant genetic correlations with other health and lifespan-related traits, while Hannum age and Horvath age did not (FIG. 7).

We then applied generalized inverse-variance weighted MR (gIVW) and MR- Egger (gEgger) on each exposure-outcome pair (FIG. lb, Method). After adjusting for multiple tests using Bonferroni correction, we discovered more than 6,000 CpG sites with significant causal effects on each trait (FIG. 1c). We then performed a pairwise conditional and colocalization (PWCoCo, Method), which is an important method to control false positive results from MR and filter out the MR signals purely driven by LD or pleiotropy ⁷⁰ . We used the conditional H4 threshold of 0.7 to identify colocalized signals and detected such signals for more than half of the CpG sites identified by MR for each trait (FIG. lb).

Since we could only perform MR and colocalization analysis on 420,509 CpG sites, the role of unmeasured CpG sites on a tested trait could not be differentiated from the measured ones. To further validate whether the effect estimated by MR can be attributed to a single CpG site, we utilized the point mutation that naturally occurs on the putative causal CpG sites (C to A or C to T), also known as meSNP. For the human methylation array, nearly 10% of CpG sites have an meSNP available. We found that the meSNPs were significantly depleted at putative causal CpG sites, suggesting that there may be a negative selection against loss-of-function mutations at these sites, possibly due to the enrichment of causal sites in regulatory regions (Figs. 8A-B). Among putative causal CpG sites with meSNPs available, we examined the correlation between the effects on the outcome trait estimated using a single meSNP and the effect estimated by MR. We observed a significant positive correlation between the two estimates (P = le-4, Pearson’s R = 0.4, Figs. 8A-B). These results suggest that the causal effect estimated by MR can be partially attributed to a single CpG site, at least in the putative causal CpG sites with available meSNPs. Yet, considering many CpG sites do not have meSNPs available and the methylation level of individual CpG site tends to be highly correlated with neighboring CpG sites ^48,71,72 , _we b _e ii _eve the putative causal CpG sites we identified also serve as tagging CpG sites for the causal regulatory region, and the causal effect size we estimated can be interpreted as the causal effect size of the tagged regulatory region.

Interestingly, the Spearman correlation of the estimated effect size of CpGs across twelve traits formed two distinct clusters, with the first cluster containing eight lifespan- and health-span-related traits, and the second all four epigenetic age measurements (FIG. Id). This observation suggests that, although all these twelve traits are genetically correlated with each other, causal CpGs do not have proportional effect sizes - the CpGs with large effects on lifespan and healthspan do not have a proportional effect size on epigenetic age measurements and vice versa.

To prioritize CpG sites with a potential causal effect on Aging-GIPl, we first filtered MR signals based on the P value threshold after Bonferroni correction. The CpG sites were then ranked according to the magnitude of the causal effect, adjusted by the colocalization probability (PP.H4). The top CpG sites whose methylation was observed to promote healthy longevity (Aging-GIPl) included cgl2122041 at the Z/ TTocus, which is associated with bone mineral density and age, cg02613937 at the TOMM40 locus, which is associated with Alzheimer’s disease and age, and cgl9047158 at the non-coding region, which is associated with gestational age and rheumatoid arthritis. The top CpG sites whose methylation was found to inhibit healthy longevity included cg04977528 at the HEYL locus, which is associated with sex and age, cg06286026 at the GRK4 locus (associated with age), cg27161488 at the C4orfl0 locus (associated with rheumatoid arthritis and age), and cgl 8744360 at the MAD1L1 locus (associated with hypotensive disorder, FIG. le). Furthermore, cgl9514613 at the APOE locus is also among the top sites that limit longevity. Genetic variants near HTT and MAML3 were also shown to significantly affect lifespan in Finnish and Japanese cohorts in a previous study ⁷³ . Both TOMM40 and APOE are known to contribute to the risk of Alzheimer’s disease and are associated with human lifespan ^74,75 . Our results suggest that the known lifespan-related effect at these loci may be mediated by DNA methylation. Moreover, we also used adjusted Aging-GIPl, where the effects on human lifespan and healthspan that are correlated with socioeconomic status are removed. We showed that after adjusting for socioeconomic status, the CpG site with the top pro-longevity effect is cg06636172 at the FOXO locus, which is a major longevity locus ^76,77 .

Example 2. Putative causal CpG sites are enriched in regulatory regions

To further understand the properties of the CpG sites identified as causal to each aging-related trait, we performed an enrichment analysis using 14 Roadmap annotations ⁷⁸ . We found that the putative causal CpGs for most traits are enriched in promoters and enhancers while depleted in quiescent regions (FIG. 2a). Furthermore, these sites were enriched in CpG shores (FIG. 9). We observed that the putative causal CpG sites for Aging-GIPl are significantly more evolutionally conserved compared to non-causal CpGs, based on both functional genomic conservation scores (Learning Evidence of Conservation from Integrated Functional genomic annotations, LECIF) and the phastCons/phyloP scores across 100 vertebrate genomes ⁷⁹ (FIG. 2b, c, FIG. 10). Moreover, the absolute value of the estimated causal effect sizes showed significant positive correlations between all three conservative scores. These results suggest that the CpG sites identified as causal for aging-related traits are more likely to be located in functional genomic elements and more evolutionarily conserved. It is well known that DNA methylation status may affect the binding of transcription factors (TFs) ⁸⁰ . To understand the relationship between putative causal CpG sites and TFs, we performed a transcription factor binding site enrichment analysis (FIG. 2d). The CpG sites causal to Aging-GIPl were significantly enriched in the binding sites of 63 TFs, including P0LR2A, ZNF24, MYC, and HDACP, while depleted in the binding sites of 19 TFs, including CTCF, CHD4, and BRD9 (FIG. 2d). In particular, P0LR2A was among the top enriched TFs in 9 of 12 traits. P0LR2A is the P0LR2 subunit (RNA polymerase II), and previous research shows that epigenetic modifications can modulate its elongation and affect alternative splicing. Our results imply that this mechanism is potentially a major contributor that mediates the effects of DNA methylation on aging ^11,12 ’ ⁸¹ . \y _e further found that there were 3 TF-binding sites (BRD4, CREB1, and F2 1) enriched with CpG sites whose methylation levels promote healthy longevity (Aging-GIPl), and 4 TF-binding sites (HDAC1, ZHXP IKZF2, a MRF ) enriched with CpG sites whose methylation levels decrease healthy longevity. BRD4 contributes to cellular senescence and promotes inflammation ³⁴ . Therefore, our findings suggest that higher DNA methylation at BRD4 binding sites may inhibit the downstream effects of BRD4 and promote healthy longevity. Similarly, previous studies showed that CREB1 is related to type II diabetes and neurodegeneration ³⁵ , and mediates the effect of calorie restriction ³⁶ . However, how DNA methylation may affect CREB 1 binding is not well studied. Our data suggest that higher methylation at CREB 1 -binding sites may promote its longevity effects. HDAC1 is a histone deacetylase, and its activity increases with aging and may promote age-related phenotypes ^30,37 . HDAC1 has been shown to specifically bind to methylated sites. Our data, therefore, support the hypothesis that HDAC1 plays a damaging role during aging, as increased DNA methylation . HDAC1 binding sites may causally inhibit healthy longevity.

Since the putative CpGs are enriched in regulatory regions and TF binding sites, we further performed a mediation analysis to investigate whether the effect top CpG hits are mediated through gene expression. The mediation effects were estimated through multivariable MR including both DNA methylation and gene expression, which dissect significant CpG-phenotype causal effects (0T) into direct effects (0D) and indirect effects through transcript levels (Method) ⁶⁵ . Among 2,255 putative causal CpGs applicable to mediation analysis for Aging-GIPl, we found 1,000 of them have their effect mediated by a major transcript (with mediation proportion > 0.03, FIG. 2e). For example, we found that the 92% of the effect of cgl 1299964 on Aging-GIPl is mediated through the expression of MAPKAP1, which is a key protein in the mTOR signalling pathway (FIG. 2e) ⁸² ; 83% of the effect of cg22120714 is mediated through the expression of KAT2A, a repressor of NF-kappa-B ⁸³ . We then performed a gene set enrichment analysis on GO and KEGG using the mediator genes for Aging-GIPl (FIG. 2f). We found that the mediators are enriched in several aging- related pathway, including mTOR signalling (P = 0.0018) and autophagosome assembly (P = 5.4e-4, FIG. 2f).

We also examined the enrichment of putative causal CpG sites in phenome- wide EWAS signals obtained from the EWAS catalog ¹² . The top enriched phenotypes included rheumatoid arthritis, HIV infection, nitrogen dioxide exposure, and maternal obesity. Interestingly, none of these conditions is primarily caused by aging. On the contrary, both rheumatoid arthritis and HIV infection are the conditions that have been suggested to accelerate aging and immunosenescence ⁸¹ . Additionally, maternal obesity is associated with accelerated metabolic aging in offspring ⁸⁴ , and nitrogen dioxide exposure is also shown to be associated with an increased risk of mortality ⁸⁵ . Among the 12 traits tested, only the putative causal CpG sites for GrimAge and Hannum age (both are epigenetic biomarker traits) were significantly enriched in the change of the CpG sites with aging, both epigenetic biomarker traits (FIG. 2e). Therefore, our results suggest that the causal CpG sites for aging are enriched in conditions that cause accelerated aging, but not in conditions that are caused by aging. This is consistent with the previous study, which suggests that differentially expressed genes reflect disease-induced rather than disease-causing changes ⁸⁶ .

Example 3. MR on epigenetic age measurements successfully recovers clock sites as putative causal CpG sites

For epigenetic age measurements, the causal CpG sites were the clock sites and the sites upstream of clock sites (FIG. 3a). To validate our EWMR approach for discovering putative causal CpG sites, we used clock sites for each clock as ground truth and investigated whether MR, when using the clock trait as outcome, could recover the clock sites as putative causal CpG sites with the correct estimated effects.

We first examined the identified putative causal CpG sites for three epigenetic age measurements with the clock models publicly available, namely HannumAge, HorvathAge, and PhenoAge ⁸ . We observed that the CpGs identified by EWMR for each epigenetic age measurement were significantly enriched with the corresponding clock sites (FIG. 3b; HannumAge P = 9.4e-9, HorvathAge P = 1.2e-12, PhenoAge P = 2.7e-6). Furthermore, EWMR predicted causal effect sizes of putative causal CpGs with the correct direction and relative magnitude; as for the three epigenetic age measurements, the estimated causal effect of MR showed a high and significant linear relationship with the actual causal effect sizes denoted by the coefficients of the clock model (FIGs. 3c-e). Notably, the enrichment and correlation we described were also robust to the choice of threshold (FIGs. 3b-e).

In MR studies, the P value is not a reliable ranking metric, as it is largely related to the number of instruments available for the exposure traits ²⁷ . As the epigenetic age GWAS provided a unique opportunity where a part of the real causal CpG sites was already known, we applied four different ranking metrics to identify an ideal ranking metric to rank putative causal CpG sites. We calculated the area under the receiver operating curve (ROC, AUROC) using the clock sites as ground truth. The AUROC measures the accuracy of binary classification, where an AUROC of 0.5 corresponds to a random classification, and an AUROC of 1 corresponds to a perfect classification. Note that since some putative causal CpGs are unknown (regulatory CpGs upstream to clock sites, FIG. 3a), the AUROC we calculated underestimated the real accuracy. However, we found that when ranking with PP-H4 weighted effect size, strikingly higher AUROCs were achieved compared to all other ranking metrics (0.99 for HannumAge, 0.83 for HorvathAge, and 0.73 for PhenoAge, FIG. 3f). As far as we know, the colocalization probability -weighted effect size has never been used for ranking MR hits. Therefore, our findings provide novel metrics that could be reliably used to prioritize MR results of molecular traits and facilitate downstream analyses.

Example 4. Existing epigenetic clocks are not enriched with CpG sites causal to aging

One open question for epigenetic clocks is whether their clock sites are causal to aging and age-related functional decline. To answer this question, we collected seven epigenetic age models in humans, namely, the Zhang clock, PhenoAge, GrimAge, PedBE, HorvathAge, HannumAge, and DunedinPACE. We then performed an enrichment analysis of putative causal CpGs for all eight lifespan/healthspan- related traits for each clock. After correcting for multiple testing, none of the existing clocks showed significant enrichment for putative causal CpGs of any of the lifespan/healthspan-related traits (FIG. 3g). PhenoAge showed a nominal significant enrichment with CpGs causal to healthspan and healthy aging, but it was not robust to the choice of thresholds. This finding suggests that, although some clocks contain CpGs causal to aging (Table 2), they, by design, favor CpG sites with a higher correlation with age and thus are not enriched with putative causal CpGs.

In contrast, even though different clocks were trained on different datasets with different methods, the causal sites identified for one clock were usually also enriched with the clock sites for other clocks, suggesting that there is a subset of CpG sites that contribute to the epigenetic age estimate of all existing epigenetic clocks, which could potentially introduce systemic bias.

Example 5. Integration of MR results and age-associated differential methylation reveals protective and deleterious epigenetic changes during aging

Another important question in epigenetic aging is the identity and number of epigenetic changes that (i) contribute to age-related damage and (ii) respond to it. We approached this question by integrating information on the causal effect and age- related differential methylation for each CpG. The protective or damaging nature of age-related differential methylation at each CpG is indicated by the product of the causal effect and age-associated differential methylation (b _age X b _MR , FIG. 4a). For example, if a higher methylation level of a certain CpG site leads to a longer lifespan or healthspan, then during aging, a decrease of the methylation level at that site would be considered as having a damaging effect, whereas an increased methylation level would be considered as having a protective effect.

The effect of DNA methylation estimated by MR is estimated through linear regression, which assumes that the relationship between DNA methylation level and lifespan-related outcome is linear. Prior to annotating protective and damaging CpGs, it is important to make sure the effect size of genetic instruments on DNA methylation levels is in the same order of magnitude as the effect of aging. We show that the effect of genetic instruments is comparable with the effect of aging by calculating the ratio between the effect of strongest cis-meQTL and age-related differential methylation for each CpG site. The median ratio was 21.8 for all significant age-associated sites and 3.9 for top 50 age-associated sites, suggesting that the median effect of genetic instruments is roughly equivalent to the effect of years of aging.

Therefore, using the age-related blood DNA methylation data estimated from 7,036 individuals (ages of 18 and 93 years, Generation Scotland cohort) ²⁷ , we separated the CpG sites causal to eight traits related to lifespan into four different categories: protective hypermethylation, deleterious hypermethylation, protective hypomethylation, and deleterious hypomethylation (FIG. 4b). Among the top 10 CpG sites whose differential methylation during aging has a relatively large impact on healthy longevity, six hypermethylated CpG sites during aging exhibit strong protective effects, including cgl8327056, cg25700533, cgl9095568, cgl7227156, cgl7113968, and cg07306253; while one hypomethylated CpG site (cg04977528) also has a protective effect. In contrast, one hypermethylated CpG site (cg26669793) and two hypomethylated CpG sites (cg25903363 and cg26628907) show damaging effects (FIG. 4b).

Contradicting the popular notion that most age-related differential methylation features are bad for the organism, our findings revealed that, in terms of the number of CpGs, there was no enrichment for either protective or damaging differential methylation during aging (FIG. 11). Note that the age-associated CpG sites are identified in cross-sectional studies, therefore, a fraction of protective sites we observed could be explained by survival bias (i.e., CpG sites that promote late-life survival). Interestingly, there is a stronger depletion of meSNP in adaptive sites compared to the damaging sites, consistent with the notion that the adaptative mechanism is stronger regulated compared to the damage (FIGs. 8A-B) ^87,88 . We also found that there was no significant correlation between the size of the causal effect and the magnitude of age-associated differential methylations (FIG. 4b, FIG. 11), suggesting that CpG sites with a greater effect on healthy longevity do not necessarily change their level of methylation during aging. This result is consistent with our findings discussed above and explains the lack of enrichment of causal sites in existing epigenetic clocks.

The product of the causal effect and age-associated differential methylation (b _age ^x b _MR ) provides an estimate of the effect of age-related differential methylation on aging-related phenotypes in a unit of time. We calculated the cumulative effect of age-associated differential methylation on Aging-GIPl by cumulative summing the effect of top 3,000 age-associated CpG sites, and calculated the empirical P-value through 10,000 permutations (FIG. 4c). Importantly, we discovered that although the number of protective and damaging CpG sites was similar, the cumulative effect of combined age-related DNA differential methylation is significantly detrimental to age-related phenotypes (P = 0.007), consistent with the overall damaging nature of aging.

Example 6. Algorithms for developing causality-informed epigenetic clocks

Although various existing epigenetic aging clock models can accurately predict the age of biological samples, they are purely based on correlation. This means that the reliability of existing clock models is highly dependent on the correlation structure of DNA methylation and phenotypes. This may result in unreliable estimates when extrapolating the model to predict the age of novel biological conditions (i.e., applying clocks to interventions that do not exist in the training population), as the correlation structure may be corrupted by the new intervention.

To overcome this problem, we developed novel epigenetic clocks that are based on putative causal CpG sites identified by EWMR (FIG. 5a). Specifically, we trained an elastic net model predicting chronological age on whole blood methylation data from 2,664 individuals ^28,29 , using CpG sites identified as causal to adjusted Aging-GIPl by EWMR (adjusted P < 0.05). In regular epigenetic clock models, the penalty weight is defined to be 1 for all CpG sites, which produces models that are purely based on correlation. Instead, we introduced a novel causality-informed elastic net model, where we assigned the feature-specific penalty factor based on the causality score for each CpG site (Method). The influence of the causality score on the feature-specific penalty factor is controlled by the causality factor T, which is an adjustable parameter. If T = 0, the whole model is reduced to a regular elastic net regression, where the penalty factor equals one for all features. When T becomes large, the model is more influenced by the causality score and tends to assign larger coefficients to the features with a higher causality score (FIG. 5 A, Methods).

Using this method, we trained the model to build the causality-informed epigenetic clock CausAge (586 sites; see Table A) using 2,664 blood samples. To separately measure adaptive and damaging DNA differential methylation during aging, we further separated putative causal CpG sites into two groups based on the causal effect size from MR and the direction of age-associated differential methylation (FIG. 4b). We then built DamAge, the damaging clock, which contains only the damaging CpG sites (1090 sites; see Table B), and AdaptAge, the protective clock, which contains only the adaptive/protective CpG sites (1000 sites, FIG. 5a; see Table C). We show that the model’s accuracy significantly decreased as the causality factor T increased (FIG. 5b, c,). This is because the causality factor T controls the trade-off between the correlation and causality score-weighted penalty factor, and the causality score is not always correlated with the predictive power of age. For example, a CpG site with a high correlation with age may not be causal to aging, and vice versa. We therefore selected causality factor T of 0.3 in the downstream analysis, which is the largest T value with MAE < 5 years in the validation set and maximized the association with mortality (FIGs. 5c-d).

Example 7. DamAge and AdaptAge clocks uncouple aging-related damage and adaptation

By design, AdaptAge contains only the CpG sites that capture protective effects against aging. Therefore, in theory, the subject predicted to be older by AdaptAge may be expected to accumulate more protective changes during aging. On the contrary, DamAge contains only the CpG sites that exhibit damaging effects, which may be considered as a biomarker of age-related damage. Therefore, we hypothesized that DamAge acceleration may be harmful and shorten life expectancy, whereas AdaptAge acceleration would be protective or neutral, which may indicate healthy longevity.

To test this hypothesis, we first analyzed the associations between human mortality and epigenetic age acceleration quantified by causality-informed clocks using 4,651 individuals from the Framingham Heart Study, FHS offspring cohort (n = 2,544 Caucasians, 54% females) and Women’s Health Initiative cohort (WHI, n = 2107 postmenopausal women, Methods). Among the three causality -informed clocks, DamAge acceleration showed the strongest positive association on mortality (P = 9.9e-12) and outperformed CausAge (P = 0.01), AdaptAge (P = 0.008), Horvath clock (P = 0.34), Hannum clock (P = 8.2e-7), and PhenoAge (P = 9.2e-l 1, FIG. 5d). This finding supports the notion that age-related damage is the main contributor to the risk of mortality, and the solely damage-base clock is better than the mixture of both damage and adaptation. In contrast, AdaptAge acceleration showed a significant negative association with mortality, suggesting that protective adaptations during aging, measured by AdaptAge, are associated with longer lifespan. In addition, epigenetic age accelerations measured by DamAge and AdaptAge were nearindependent (Pearson's R = 0.14, FIG. 13). These findings highlight the importance of separating adaptive and damaging age-associated differential methylation when building aging clock models.

Interestingly, although the clock accuracy monotonically decreased as the causality factor T increased, the association between mortality and epigenetic age acceleration did not follow the same trend (FIG. 5d). Especially for DamAge, the mortality association increased as the T increased and peaked when T was around 0.3. Also, DamAge consistently outperformed CausAge in predicting mortality risk, even though CausAge was more accurate in age prediction (FIGs. 5b-e), the association between CausAge and mortality may be weakened due to the inclusion of adaptive sites. This suggests that although the introduction of the causality score and separation of damaging CpGs may decrease the accuracy of the clock in terms of predicting chronological age, it improves the prediction of aging-related phenotypes.

Induced pluripotent stem cell (iPSC) reprogramming is one of the most robust rejuvenation models, which was shown to be able to strongly reverse the epigenetic age of cells ^11,30 . We applied the causality-informed clock models to reprogramming of fibroblasts to iPSC ³⁶ . For comparison, we also included five published epigenetic models, namely Horvath Age, Hannum Age, Pheno Age, Grim Age and DunedinPACE. The Horvath and Hannum clocks were trained on chronological age ^38,39 , PhenoAge was trained on the age-adjusted by health-related phenotypes ^40,41 , GrimAge was trained on mortality ⁸⁹ , and DunedinPACE was trained to predict the pace of aging ⁴⁰ . Consistent with Horvath clock, Hannum clock, PhenoAge, and GrimAge, DamAge revealed that epigenetic age decreased during iPSC reprogramming, but with a stronger negative correlation with the time of reprogramming and higher statistical significance (R = -0.93, P = 4e-12, FIG. 5f). This observation suggests that DamAge may better capture the damage-removal effect of iPSC reprogramming. On the contrary, AdaptAge increased significantly during the reprogramming process (R = 0.86, P = 1.3e-8), suggesting that protective age- associated differential methylation does not capture the rejuvenation effect and that in fact cells may acquire even more protective changes during iPSC reprogramming. Example 8. Causality-informed epigenetic clocks capture damage and aging-related effects

To further examine how Dam Age and AdaptAge capture age-related damage and protective adaptations, respectively, we tested performance of causality-informed clocks using various datasets. For comparison, we included two 1st generation clocks (Horvath age and Hannum age), which are trained solely on chronological age, and three 2nd generation clocks (DunedinPACE, PhenoAge, and GrimAge), which are trained on mortality- and health-related outcomes.

We first examined several aging-related conditions, namely atherosclerosis, cancer, and hypertension (FIG. 6a). We analyzed blood samples from clinical atherosclerosis patients (n = 8) and healthy donors (n = 8) in the LVAD study ⁹⁰ . All eight clocks tested showed that the atherosclerosis patients were significantly biologically older than healthy controls (FIG. 6a). We also analyzed 70 prostate cancer cases with good or poor prognosis ⁹¹ . Only DamAge successfully detected a significant age acceleration in patients with bad cancer prognosis (P = 0.039), while Hannum age detected a significant inverse effect where the patients with good prognosis were age accelerated (P = 0.044). For hypertensive heart disease, we analyzed blood samples from 44 hypertensive patients and 44 healthy controls ⁹² . Both CausAge and DamAge showed significant age acceleration in hypertensive patients (CausAge P = 0.002, DamAge P = 0.04). Similar effects could be detected with GrimAge (P = 0.002) and DunedinPACE (P = 0.02), but not with AdaptAge, PhenoAge, and two 1st generation clocks. These results suggest that DamAge could more robustly represent the effect of age-related conditions, compared to the published 1st and 2nd generation clocks.

Next, we examined conditions that specifically promote age-related damage (FIG. 6b). Smoking is a well-known risk factor for many age-related diseases, and it also causes DNA damage and oxidative stress ⁹³ . We compared the epigenetic age of smokers (n = 40) and non-smokers (n = 40) ⁹⁴ . CausAge (P = 0.004) and DamAge (P = 0.006), together with all three 2nd generation clocks could detect significant ageacceleration among smokers, while AdaptAge and two 1st generation clocks did not. Progeroid syndrome is a group of rare genetic disorders that cause premature aging ⁹⁵ . We analyzed blood cell samples from healthy donors (n = 3), and patients with Hutchinson-Gilford Progeria Syndrome (HGP, n = 3) and Werner Syndrome (n = 4) ⁹⁶ . We observed significant DamAge acceleration in both HGP (P = 0.004) and Werner Syndrome (P = 5e-4) compared to healthy controls. Similar effects were detected also with PhenoAge and GrimAge. Hannum age and DunedinPACE detected age acceleration in Werner Syndrome but not in HGP, while no significant effect was found by other clocks (FIG. 6b). We then analyzed dermis and epidermis samples with or without sun exposure (n = 10 per group) in older adults (age > 60) ⁹⁷ . As the exposure to ultraviolet promotes DNA damage and aging, it may be considered a model of age-related damage. As expected, we observed significant DamAge acceleration in sun-exposed epidermis compared to sun-protected epidermis (P = 2e- 5), while no significant effect was observed in the dermis tissue. AdaptAge of the sun- exposed epidermis was significantly lower (P = 0.01). Surprisingly, based on most other published clocks (including Horvath age, Hannum age, and DunedinPACE), the sun-exposed epidermis was predicted to be significantly younger than sun-protected epidermis. Only GrimAge showed the expected effect direction but did not reach statistical significance (P = 0.1).

Paraoxonase 1 (PON1) is one of most studied genes associated with cardiovascular disease, oxidative stress, inflammation, and healthy aging ⁴² . Specifically, PON1 plays an important role in detoxifying organophosphorus compounds and removing harmful oxidized lipids ⁷ . The genetic variant of PON1 (R192Q) significantly decreases PON1 activity and is known to be associated with an increased risk of cardiovascular disease and neurodegenerative diseases ⁴³ . Interestingly, the PON1 Q allele is significantly depleted in centenarians ⁴⁴ . We analyzed the relationship between PON1 activity and epigenetic age in 48 whole blood samples (FIG. 6a) ⁹⁸ . DamAge shows a significant negative correlation with PON1 activity (R = -0.55, p = 0.0062), whereas AdaptAge showed a significant positive correlation with /Y>W/ activity (R = 0.69, p = 0.0003). Again, this association was not observed by other epigenetic clocks, except for Horvath age, but with a less significant negative correlation (P = 0.04). Thus DamAge can reliably detect damage- related biological age acceleration.

Causality-informed clocks could also capture the aging-related effects of short-term interventions. We first examined the effect of human umbilical cord plasma concentrate injection, which was reported to have age reversal effects ⁹⁹ . In this study, 18 elderly participants were treated with human umbilical cord plasma concentrate injection weekly (1 ml intramuscular) over a 10-week period. We found that this rejuvenation effect could only be captured with DamAge (P = 0.04) and GrimAge (P = 0.04), but not with other clocks (FIG. 6c). Similarly, a 6-week omega-3 fatty acid supplementation in overweight subjects (n = 34) ¹⁰ °, which was shown to be protective against age-related cardiovascular diseases, significantly increased AdaptAge (P = 0.009) and reduced Dam Age (P = 0.02, FIG. 6c). We also found that short-term treatment with cigarette smoke condensate in bronchial epithelial cells significantly accelerated DamAge (P = 0.002) but did not affect other tested clocks (FIG. 6c). Together, our data demonstrate the importance of separating damage and adaptation when building biomarkers of aging and provide novel tools to quantify aging and rejuvenation.

Previous studies have shown that anti-aging interventions delivered during development could prolong lifespan and healthspan, including calorie restriction (CR) ¹⁰¹ and rapamycin treatment ¹⁰² . Small for gestational age (SGA) is a condition defined as birth weight less than the 10th percentile for gestational age ¹⁰³ . We found that children with SGA have a significantly lower DamAge and higher AdaptAge than children with normal birth weight. These effects were not captured by other epigenetic clocks tested. SGA is usually considered a pathological condition; some studies suggest that it may be because early life benefits can be reversed in later life by exposure to excess nutrients ¹⁰⁴ . The different roles of SGA in the early and late stages of life may need to be further investigated in future studies.

In vitro fertilization (IVF) is a common method of treating infertility. Yet, previous studies have shown that IVF may increase the risk of perinatal morbidity and mortality ¹⁰⁵ . We analyzed the DNA methylation data from neonatal blood spots of 137 newborns conceived unassisted (NAT), through intrauterine insemination (IUI), or through IVF using fresh or cryopreserved (frozen) embryo transfer ¹⁰⁶ . We found that IVF-conceived newborns using fresh or cryopreserved embryos had higher DamAge acceleration and lower AdaptAge than NAT-conceived newborns. On the other hand, lUI-conceived newborns showed no differences in their DamAge and AdaptAge with controls. This effect could not be observed by other five epigenetic clocks tested, except for Horvath age. Genomic imprinting is an epigenetic mechanism that controls the expression of parent-of-origin-dependent gene, which plays an important role in embryonic development and has a lifelong impact on health ¹⁰⁷ . Some imprinting genes are known to be associated with metabolic disorders and aging (e.g., IGF2-H19) ^{108 109} \y _e analyzed peripheral blood DNA methylation data from patients with single-locus or Multi-loci imprinting disturbances (SLID or MLID), which is the condition of losing methylation at single or multiple imprinting centers ¹¹⁰ . Similar to IVF, we found that patients with imprinting disorders showed significantly higher DamAge and lower AdaptAge. Together, these results suggest that DamAge and AdaptAge may serve as preferred biomarkers for the events affecting aging traits during development.

Table 1. Datasets used in this study

Dataset Description

GWAS data meQTLs were obtained from the Genetics of DNA Methylation Consortium (GoDMC). DNA meQTLs methylation levels were measured in whole blood samples from 36 cohorts, including 27,750

European subjects. 420,509 CpG sites were analyzed (Min et al.) ⁴⁵ .

First genetic principal component of six human aging traits - healthspan, father and mother Aging-GIP1 lifespan, exceptional longevity, frailty index and self-rated health, which captures both length of life and indices of mental and physical wellbeing (Timmers et al.) ⁵³ .

Aging-GIP1-Aging-GIP1 adjusted for household income and socioeconomic deprivation, from the same adj GWAS study as above ⁵³ . aging ^hy The ^mu ' ^t ' ^var ' ^ate g ^enom i ^{c scan} of healthspan, lifespan, and longevity (Timmers et al.) ⁵⁴ .

. GWAS of lifespan from 512,047 mothers and 500,193 fathers of European ancestry (Timmers et

Lirespan _g | ^ ₄₈

11 ,262 subjects of European ancestry with a lifespan above the 90th percentile as the case Longevity group and 25,483 control subjects whose age at the last visit was below the 60th percentile age (Deelen et al.) ⁴⁹ .

The age of the first incidence of any major age-related disease, including dementia, congestive _H ... heart failure, diabetes, chronic obstructive pulmonary disease, stroke, cancer, myocardial ea span j _n f _arc t j _{on as we} || _as the incidence of death. The GWAS of healthspan included 300,447 subjects of European ancestry from the UK Biobank cohort (Zenin et al.) ⁵⁰ .

_ ... Calculated based on the cumulative number of health deficits during aging. The frailty index index^ GWAS included 164,610 UK Biobank participants aged 60-70 years and 10,616 Swedish

TwinGene participants aged 41-87 years (Atkins et al.) ⁵⁵ .

Self-rated Self-rated health GWAS was based on questionnaire responses on a scale of 6-5 in UK Biobank health cohort, downloaded from Pan-UKBB project.

Horvath a _e ^st 9 ^enerat ' ^on multi-tissue clock trained on chronological age, GWAS was performed on 34,710 o ^{rva a} ^ ^e European ancestry and 6,195 African American individuals (McCartney et al) ⁵²

Hannum 1st generation blood clock trained on chronological age, from the same GWAS study as above age ⁵² .

PhenoAge 2st generatjon bipod clock trained on phenotypic age, from the same; G\^S study as above ⁵² . GrimAge 2st generation blood clock trained on mortality risk, from the same GWAS study as above ⁵² .

. GEO data .

This study conducted DNA methylation analyses of blood samples from atherosclerosis patients and healthy donors.

GSE127985 DNA methylation changes in prostate cancer cases and it's prognosis.

This study analyzed peripheral whole blood DNA methylation profiles of pregnant women at

GSE192918 different stages of gestation and post-delivery, identifying changes in DNA methylation patterns associated with different time points during pregnancy.

GSE193795 ^ ^enome-w '^ ^e DNA methylation profiling was performed on 44 hypertensive and 44 healthy

° control samples, revealing distinct DNA methylation patterns associated with hypertension.

GSE210245 ^{DNA meth} y ^|ation data f ^rom human whole blood samples were analyzed to assess the impact of ° treatment with human umbilical cord plasma concentrate injection.

GSE51954 Genome-wide DNA methylation profiling was conducted on epidermal and dermal samples obtained from sun-exposed and sun-protected body sites. This study compared global methylation changes in buccal cells between smokers, moist snuff

GSE94876 consumers, a^nd non-tobacco consumers.

This study aimed to explore genome-wide DNA methylation changes and identify altered

GSE98056 biological pathways resulting from n-3 fatty acid supplementation in overweight and obese individuals.

GSE101673 DNA methylation data for cigarette smoke condensate treated celL

GSE78773 ^' ^{s stu} ^y identified multi-locus methylation disturbances in individuals with different methylation patterns, including patients with Temple and Angelman syndromes

GSE90117 This study identified the relationship between PON 1 activity, allele, and DNA methylation.

GSE79257 h' ^{s stu} ^y analyzed DNA methylation in infants born through different assisted reproductive techniques and unassisted conception, utilizing archived Guthrie cards for methylation profiling. This study analyzed DNA methylation in B cells from patients with Hutchinson-Gilford Progeria

GSE42865 Syndrome (HGP) and Werner Syndrome and controls.

Table 2. Putative causal CpG sites in existing epigenetic clocks

P, Protective; D, deleterious References

1. Chauhan, A. et al. Systems biology approaches in aging research. Interdiscip Top Gerontol 40, 155-76 (2015).

2. Sen, P., Shah, P. P., Nativio, R. & Berger, S. L. Epigenetic Mechanisms of Longevity and Aging. Cell 166, 822-839 (2016).

3. Guibert, S. & Weber, M. Functions of DNA methylation and hydroxymethylation in mammalian development. Curr Top Dev Biol 104, 47-83 (2013).

4. Jones, P. A. Functions of DNA methylation: islands, start sites, gene bodies and beyond. Nat. Rev. Genet. 13, 484-492 (2012).

5. Moqri, M. et al. PRC2 clock: a universal epigenetic biomarker of aging and rejuvenation. 2022.06.03.494609 Preprint at https://doi.org/10.1101/2022.06.03.494609 (2022).

6. Kane, A. E. & Sinclair, D. A. Epigenetic changes during aging and their reprogramming potential. Crit Rev Biochem Mol Biol 54, 61-83 (2019).

7. Lopez-Otin, C., Blasco, M. A., Partridge, L., Serrano, M. & Kroemer, G. The Hallmarks of Aging. Cell 153, 1194-1217 (2013).

8. Bell, C. G. et al. DNA methylation aging clocks: challenges and recommendations. Genome Biol 20, 249 (2019).

9. A. Reynolds, C. et al. A decade of epigenetic change in aging twins: Genetic and environmental contributions to longitudinal DNA methylation. Aging Cell 19, (2020).

10. Field, A. E. et al. DNA Methylation Clocks in Aging: Categories, Causes, and Consequences. Mol Cell 71, 882-895 (2018).

11. Horvath, S. DNA methylation age of human tissues and cell types. Genome Biol. 14, R115 (2013).

12. Hannum, G. et al. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol. Cell 49, 359-367 (2013).

13. Chen, B. H. et al. DNA methylation-based measures of biological age: metaanalysis predicting time to death. Aging 8, 1844-1865 (2016).

14. Marioni, R. E. et al. The epigenetic clock is correlated with physical and cognitive fitness in the Lothian Birth Cohort 1936. Int. J. Epidemiol. 44, 1388— 1396 (2015). 15. Quach, A. et al. Epigenetic clock analysis of diet, exercise, education, and lifestyle factors. Aging 9, 419-446 (2017).

16. Rutledge, J., Oh, H. & Wyss-Coray, T. Measuring biological age using omics data. Nat. Rev. Genet. (2022) doi : 10.1038/s41576-022-00511 -7.

17. Kapur, K. Chapter 14 - Principles of Biostatistics, in Clinical and Translational Science (Second Edition) (eds. Robertson, D. & Williams, G. H.) 243-260 (Academic Press, 2017). doi: 10.1016/B978-0-12-802101-9.00014-4.

18. Davey Smith, G. & Hemani, G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet 23, R89-98 (2014).

19. Evans, D. M. & Davey Smith, G. Mendelian Randomization: New Applications in the Coming Age of Hypothesis-Free Causality. Annu Rev Genomics Hum Genet 16, 327-50 (2015). 0. Neumeyer, S., Hemani, G. & Zeggini, E. Strengthening Causal Inference for Complex Disease Using Molecular Quantitative Trait Loci. Trends Mol Med 26, 232-241 (2020). 1. Vosa, U. et al. Large-scale cis- and trans-eQTL analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression. Nat.

Genet. 53, 1300-1310 (2021). 2. Garrido-Martin, D., Borsari, B., Calvo, M., Reverter, F. & Guigo, R. Identification and analysis of splicing quantitative trait loci across multiple tissues in the human genome. Nat. Commun. 12, 727 (2021). 3. He, B., Shi, J., Wang, X., Jiang, H. & Zhu, H.-J. Genome-wide pQTL analysis of protein expression regulatory networks in the human liver. BMC Biol. 18, 97 (2020). 4. Kraus, W. E. et al. Metabolomic Quantitative Trait Loci (mQTL) Mapping Implicates the Ubiquitin Proteasome System in Cardiovascular Disease Pathogenesis. PLoS Genet 11, el005553 (2015). 5. Huan, T. et al. Genome-wide identification of DNA methylation QTLs in whole blood highlights pathways for cardiovascular disease. Nat Commun 10, 4267 (2019). 6. Richardson, T. G. et al. Systematic Mendelian randomization framework elucidates hundreds of CpG sites which may mediate the influence of genetic variants on disease. Hum. Mol. Genet. 27, 3293-3304 (2018). 27. McCartney, D. L. et al. An epigenome-wide association study of sex-specific chronological ageing. Genome Med. 12, 1 (2019).

28. Huh, I., Zeng, J., Park, T. & Yi, S. V. DNA methylation and transcriptional noise. Epigenetics Chromatin 6, 9 (2013).

29. Kim, S., Park, H. J., Cui, X. & Zhi, D. Collective effects of long-range DNA methylations predict gene expressions and estimate phenotypes in cancer. Sci. Rep. 10, 1-12 (2020).

30. Lu, Y. et al. Reprogramming to recover youthful epigenetic information and restore vision. Nature 588, 124-129 (2020).

31. Porcu, E. et al. Differentially expressed genes reflect disease-induced rather than disease-causing changes in the transcriptome. Nat. Commun. 12, 5647 (2021).

32. Emdin, C. A., Khera, A. V. & Kathiresan, S. Mendelian Randomization. JAMA 318, 1925-1926 (2017).

33. Lin, D. et al. Characterization of cross-tissue genetic-epigenetic effects and their patterns in schizophrenia. Genome Med. 10, 13 (2018).

34. Burgess, S., Zuber, V., Valdes-Marquez, E., Sun, B. B. & Hopewell, J. C. Mendelian randomization with fine-mapped genetic data: Choosing from large numbers of correlated instrumental variables. Genet. Epidemiol. 41, 714-725 (2017).

35. Lehne, B. et al. A coherent approach for analysis of the Illumina HumanMethylation450 BeadChip improves data quality and performance in epigenome-wide association studies. Genome Biol. 16, 37 (2015).

36. Ohnuki, M. et al. Dynamic regulation of human endogenous retroviruses mediates factor-induced reprogramming and differentiation potential. Proc. Natl. Acad. Sci. U. S. A. Ill, 12426-12431 (2014).

37. Rando, T. A. & Chang, H. Y. Aging, Rejuvenation, and Epigenetic Reprogramming: Resetting the Aging Clock. Cell 148, 46-57 (2012).

38. Tollefsbol, T. O. Techniques for Analysis of Biological Aging. Methods Mol. Biol. Clifton NJ 371, 1-7 (2007).

39. Tosato, M., Zamboni, V., Ferrini, A. & Cesari, M. The aging process and potential interventions to extend life expectancy. Clin. Interv. Aging 2, 401-412 (2007).

40. Fontana, L., Partridge, L. & Longo, V. D. Extending healthy life span— from yeast to humans. Science 328, 321-6 (2010). 41. Milman, S. et al. Low insulin-like growth factor-1 level predicts survival in humans with exceptional longevity. Aging Cell 13, 769-771 (2014).

42. Schumacher, B. et al. Delayed and Accelerated Aging Share Common Longevity Assurance Mechanisms. PLOS Genet. 4, el000161 (2008).

43. Walther, D. M. et al. Widespread Proteome Remodeling and Aggregation in Aging C. elegans. Cell 161, 919-932 (2015).

44. Goto, K. et al. Protective mechanism against age-associated changes in the peripheral nerves. Life Sci. 253, 117744 (2020).

45. Min, J. L. et al. Genomic and phenotypic insights from an atlas of genetic effects on DNA methylation. Nat. Genet. 53, 1311-1321 (2021).

46. Griffith, G. et al. Collider bias undermines our understanding of COVID-19 disease risk and severity. medRxiv 2020.05.04.20090506 (2020) doi: 10.1101/2020.05.04.20090506.

47. Zhou, Z., Rahme, E., Abrahamowicz, M. & Pilote, L. Survival Bias Associated with Time-to-Treatment Initiation in Drug Effectiveness Evaluation: A Comparison of Methods. Am. J. Epidemiol. 162, 1016-1023 (2005).

48. Timmers, P. R. et al. Genomics of 1 million parent lifespans implicates novel pathways and common diseases and distinguishes survival chances. eLife 8, e39856 (2019).

49. Deelen, J. et al. A meta-analysis of genome-wide association studies identifies multiple longevity genes. Nat. Commun. 10, 3669 (2019).

50. Zenin, A. et al. Identification of 12 genetic loci associated with human healthspan. Commun. Biol. 2, 41 (2019).

51. Kojima, G., Iliffe, S. & Walters, K. Frailty index as a predictor of mortality: a systematic review and meta-analysis. Age Ageing 47, 193-200 (2018).

52. McCartney, D. L. et al. Genome-wide association studies identify 137 genetic loci for DNA methylation biomarkers of aging. Genome Biol. 22, 194 (2021).

53. Timmers, P. R. H. J. et al. Mendelian randomization of genetically independent aging phenotypes identifies LPA and VCAM1 as biological targets for human aging. Nat. Aging !, 19-30 (2022).

54. Timmers, P., Wilson, J. F., Joshi, P. K. & Deelen, J. Multivariate genomic scan implicates novel loci and haem metabolism in human ageing. Nat Commun 11, 3570 (2020). 55. Atkins, J. L. et al. A genome-wide association study of the frailty index highlights brain pathways in ageing. Aging Cell 20, el3459 (2021).

56. Bulik-Sullivan, B. K. et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet 47, 291-5 (2015).

57. Burgess, S., Butterworth, A. & Thompson, S. G. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet. Epidemiol. 37, 658-665 (2013).

58. Zhao, H. et al. Proteome-wide Mendelian randomization in global biobank metaanalysis reveals multi-ancestry drug targets for common diseases.

2022.01.09.21268473 Preprint at https://doi.Org/10.l 101/2022.01.09.21268473 (2022).

59. Yavorska, O. O. & Burgess, S. MendelianRandomization: an R package for performing Mendelian randomization analyses using summarized data. Int. J. Epidemiol. 46, 1734-1739 (2017).

60. Hemani, G. et al. The MR-Base platform supports systematic causal inference across the human phenome. eLife 7, e34408 (2018).

61. Bowden, J., Davey Smith, G. & Burgess, S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol 44, 512-25 (2015).

62. Hemani, G., Tilling, K. & Davey Smith, G. Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet.

13, el007081 (2017).

63. Wallace, C. Eliciting priors and relaxing the single causal variant assumption in colocalisation analyses. PLOS Genet. 16, el008720 (2020).

64. Zheng, J. et al. Phenome-wide Mendelian randomization mapping the influence of the plasma proteome on complex diseases. Nat. Genet. 52, 1122-1131 (2020).

65. Sadler, M. C., Auwerx, C., Lepik, K., Porcu, E. & Kutalik, Z. Quantifying the role of transcript levels in mediating DNA methylation effects on complex traits and diseases. Nat. Commun. 13, 7559 (2022).

66. Zou, H. & Hastie, T. Regularization and variable selection via the elastic net. J. R. Stat. Soc. Ser. B Stat. Methodol. 67, 301-320 (2005).

67. Handl, L., Jalali, A., Scherer, M., Eggeling, R. & Pfeifer, N. Weighted elastic net for unsupervised domain adaptation with application to age prediction from DNA methylation data. Bioinformatics 35, i 154— i 163 (2019). 68. Tay, J. K., Aghaeepour, N., Hastie, T. & Tibshirani, R. Feature-weighted elastic net: using ‘features of features’ for better prediction. ArXiv200601395 Cs Stat (2020).

69. Kundaje, A. et al. Integrative analysis of 111 reference human epigenomes. Nature 518, 317-330 (2015).

70. Robinson, J. W. et al. An efficient and robust tool for colocalisation: Pair-wise Conditional and Colocalisation (PWCoCo). 2022.08.08.503158 Preprint at https://doi.org/10.1101/2022.08.08.503158 (2022).

71. Shao, Y. et al. DNA methylation of TOMM40-APOE-APOC2 in Alzheimer’s disease. J. Hum. Genet. 63, 459-471 (2018).

72. Goh, L. K. et al. TOMM40 alterations in Alzheimer’s disease over a 2-year follow-up period. J. Alzheimer s Dis. J AD 44, 57-61 (2015).

73. Hu, S. et al. DNA methylation presents distinct binding sites for human transcription factors. eLife 2, e00726 (2013).

74. Gonzalez-Rodriguez, P., Klionsky, D. J. & Joseph, B. Autophagy regulation by RNA alternative splicing and implications in human diseases. Nat. Commun. 13, 2735 (2022).

75. Saldi, T., Cortazar, M. A., Sheridan, R. M. & Bentley, D. L. Coupling of RNA Polymerase II Transcription Elongation with Pre-mRNA Splicing. J. Mol. Biol. 428, 2623-2635 (2016).

76. Hua, T. et al. BRD4 Inhibition Attenuates Inflammatory Pain by Ameliorating NLRP3 Inflammasome-Induced Pyroptosis. Front. Immunol. 13, 837977 (2022).

77. Wang, H. et al. BRD4 contributes to LPS-induced macrophage senescence and promotes progression of atherosclerosis-associated lipid uptake. Aging 12, 9240- 9259 (2020).

78. Herzig, S. et al. CREB regulates hepatic gluconeogenesis through the coactivator PGC-1. Nature 413, 179-183 (2001).

79. Fusco, S. et al. A role for neuronal cAMP responsive-element binding (CREB)-l in brain responses to calorie restriction. Proc. Natl. Acad. Sci. 109, 621-626 (2012).

80. Willis-Martinez, D., Richards, H. W., Timchenko, N. A. & Medrano, E. E. Role of HD AC 1 in senescence, aging, and cancer. Exp. Gerontol. 45, 279-285 (2010).

81. Levine, M. E. et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging 10, 573-591 (2018). 82. Yuan, Y. et al. Characterization of Sinl Isoforms Reveals an mTOR-Dependent and Independent Function of Sinly. PLOS ONE 10, e0135017 (2015).

83. Mao, X. et al. GCN5 is a required cofactor for a ubiquitin ligase that targets NF- kappaB/RelA. Genes Dev. 23, 849-861 (2009).

84. Belsky, D. W. et al. Quantification of the pace of biological aging in humans through a blood test, the DunedinPoAm DNA methylation algorithm. eLife 9, e54870 (2020).

85. Stewart, E. A. et al. Uterine fibroids. Nat. Rev. Dis. Primer 2, 16043 (2016).

86. George, J. W. et al. Integrated Epigenome, Exome, and Transcriptome Analyses Reveal Molecular Subtypes and Homeotic Transformation in Uterine Fibroids. Cell Rep. 29, 4069-4085. e6 (2019).

87. Aledo, J. C. & Blanco, J. M. Aging is Neither a Failure nor an Achievement of Natural Selection. Curr. Aging Sci. 8, 4-10.

88. Flatt, T. & Partridge, L. Horizons in the evolution of aging. BMC Biol 16, 93 (2018).

89. Lu, A. T. et al. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging 11, 303-327 (2019).

90. Istas, G. et al. Identification of differentially methylated BRCA1 and CRISP2 DNA regions as blood surrogate markers for cardiovascular disease. Sci. Rep. 7, 5120 (2017).

91. Toth, R. et al. Random forest-based modelling to detect biomarkers for prostate cancer progression. Clin. Epigenetics 11, 148 (2019).

92. Bai, C. et al. Oviductal Glycoprotein 1 Promotes Hypertension by Inducing Vascular Remodeling Through an Interaction With MYH9. Circulation 146, 1367-1382 (2022).

93. Csiszar, A. et al. Oxidative stress and accelerated vascular aging: implications for cigarette smoking. Front. Biosci. Landmark Ed. 14, 3128-3144 (2009).

94. Jessen, W. J., Borgerding, M. F. & Prasad, G. L. Global methylation profiles in buccal cells of long-term smokers and moist snuff consumers. Biomark. Biochem. Indic. Expo. Response Susceptibility Chem. 23, 625-639 (2018).

95. Kudlow, B. A., Kennedy, B. K. & Monnat, R. J. Werner and Hutchinson- Gilford progeria syndromes: mechanistic basis of human progeroid diseases. Nat. Rev. Mol. Cell Biol. 8, 394-404 (2007). 96. Heyn, H., Moran, S. & Esteller, M. Aberrant DNA methylation profiles in the premature aging disorders Hutchinson-Gilford Progeria and Werner syndrome. Epigenetics 8, 28-33 (2013).

97. Vandiver, A. R. et al. Age and sun exposure-related widespread genomic blocks of hypomethylation in nonmalignant skin. Genome Biol. 16, 80 (2015).

98. Declerck, K. et al. Interaction between prenatal pesticide exposure and a common polymorphism in the PON1 gene on DNA methylation in genes associated with cardio-metabolic disease risk-an exploratory study. Clin. Epigenetics 9, 35 (2017).

99. Clement, J. et al. Umbilical cord plasma concentrate has beneficial effects on DNA methylation GrimAge and human clinical biomarkers. Aging Cell 21, el3696 (2022).

100. Tremblay, B. L. et al. Epigenetic changes in blood leukocytes following an omega-3 fatty acid supplementation. Clin. Epigenetics 9, 43 (2017).

101. Sun, L., Sadighi Akha, A. A., Miller, R. A. & Harper, J. M. Life-span extension in mice by preweaning food restriction and by methionine restriction in middle age. J. Gerontol. A. Biol. Sci. Med. Sci. 64, 711-722 (2009).

102. Shindyapina, A. V. et al. Rapamycin treatment during development extends life span and health span of male mice and Daphnia magna. Sci. Adv. 8, eabo5482 (2022).

103. Osuchukwu, O. O. & Reed, D. J. Small for Gestational Age. in StatPearls (StatPearls Publishing, 2022).

104. Ribeiro, A. M., Lima, M. de C., de Lira, P. I. C. & da Silva, G. A. P. Low birth weight and obesity: causal or casual casual association? Rev. Paul. Pediatr. 33, 340-348 (2015).

105. Sabban, H., Zakhari, A., Patenaude, V., Tulandi, T. & Abenhaim, H. A. Obstetrical and perinatal morbidity and mortality among in-vitro fertilization pregnancies: a population-based study. Arch. Gynecol. Obstet. 296, 107-113 (2017).

106. Estill, M. S. et al. Assisted reproductive technology alters deoxyribonucleic acid methylation profiles in bloodspots of newborn infants. Fertil. Steril. 106, 629- 639. elO (2016). 107. Monk, D., Mackay, D. J. G., Eggermann, T., Maher, E. R. & Riccio, A. Genomic imprinting disorders: lessons on how genome, epigenome and environment interact. Nat. Rev. Genet. 20, 235-248 (2019).

108. Ratajczak, M. Z. Igf2-H19, an imprinted tandem gene, is an important regulator of embryonic development, a guardian of proliferation of adult pluripotent stem cells, a regulator of longevity, and a ‘passkey’ to cancerogenesis. Folia Histochem. CytobioL 50, 171-179 (2012).

109. Butler, M. G. Genomic imprinting disorders in humans: a mini-review. J. Assist. Reprod. Genet. 26, 477 (2009).

110. Bens, S. et al. Phenotypic spectrum and extent of DNA methylation defects associated with multilocus imprinting disturbances. Epigenomics 8, 801-816 (2016).

OTHER EMBODIMENTS

It is to be understood that while the invention has been described in conjunction with the detailed description thereof, the foregoing description is intended to illustrate and not limit the scope of the invention, which is defined by the scope of the appended claims. Other aspects, advantages, and modifications are within the scope of the following claims.