Improved precision of epigenetic clock estimates across tissues and its implication for biological ageing.

Zhang Q, Vallerga CL, Walker RM, Lin T, Henders AK, Montgomery GW, He J, Fan D, Fowdar J, Kennedy M, Pitcher T, Pearson J, Halliday G, Kwok JB, Hickie I, Lewis S, Anderson T, Silburn PA, Mellick GD, Harris SE, Redmond P, Murray AD, Porteous DJ, Haley CS, Evans KL, McIntosh AM, Yang J, Gratten J, Marioni RE, Wray NR, Deary IJ, McRae AF, Visscher PM. · 2019

Read Improved precision of epigenetic clock estimates across tissues and its implication for biological ageing. by Zhang Q, Vallerga CL, Walker RM, Lin T, Henders AK, Montgomery GW, He J, Fan D, Fowdar J, Kennedy M, Pitcher T, Pearson J, Halliday G, Kwok JB, Hickie I, Lewis S, Anderson T, Silburn PA, Mellick GD, Harris SE, Redmond P, Murray AD, Porteous DJ, Haley CS, Evans KL, McIntosh AM, Yang J, Gratten J, Marioni RE, Wray NR, Deary IJ, McRae AF, Visscher PM. (2019) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc6708158.txt.

Introduction

Zhang and co-authors tested how to make epigenetic clocks more exact. These clocks guess a person's age from chemical marks on DNA. A gap between clock age and real age has been linked to death risk. The team asked whether a sharper clock changes that link.

They trained age predictors on as many as 13,661 DNA samples, mostly from blood. As training data grew, the clock got far more exact. With the best clock, the age gap no longer linked to death in two Scottish birth cohorts. Small training sets, they warn, can be thrown off by the mix of cell types in a sample.

Catalog record

CREATORZhang Q, Vallerga CL, Walker RM, Lin T, Henders AK, Montgomery GW, He J, Fan D, Fowdar J, Kennedy M, Pitcher T, Pearson J, Halliday G, Kwok JB, Hickie I, Lewis S, Anderson T, Silburn PA, Mellick GD, Harris SE, Redmond P, Murray AD, Porteous DJ, Haley CS, Evans KL, McIntosh AM, Yang J, Gratten J, Marioni RE, Wray NR, Deary IJ, McRae AF, Visscher PM.
YEAR2019
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC6708158

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Improved precision of epigenetic clock estimates across tissues and its implication for biological ageing By Qian Zhang, Costanza L Vallerga, Rosie M Walker, Tian Lin, Anjali K Henders, Grant W Montgomery, Ji He, Dongsheng Fan, Javed Fowdar, Martin Kennedy, Toni Pitcher, John Pearson Published in: Genome Medicine DOI: 10.1186/s13073-019-0667-1 ABSTRACT Abstract Background DNA methylation changes with age. Chronological age predictors built from DNA methylation are termed ‘epigenetic clocks’. The deviation of predicted age from the actual age (‘age acceleration residual’, AAR) has been reported to be associated with death. However, it is currently unclear how a better prediction of chronological age affects such association. Methods In this study, we build multiple predictors based on training DNA methylation samples selected from 13,661 samples (13,402 from blood and 259 from saliva). We use the Lothian Birth Cohorts of 1921 (LBC1921) and 1936 (LBC1936) to examine whether the association between AAR (from these predictors) and death is affected by (1) improving prediction accuracy of an age predictor as its training sample size increases (from 335 to 12,710) and (2) additionally correcting for confounders (i.e., cellular compositions). In addition, we investigated the performance of our predictor in non-blood tissues. Results We found that in principle, a near-perfect age predictor could be developed when the training sample size is sufficiently large. The association between AAR and mortality attenuates as prediction accuracy increases. AAR from our best predictor (based on Elastic Net, https://github.com/qzhang314/DNAm-based-age-predictor) exhibits no association with mortality in both LBC1921 (hazard ratio = 1.08, 95% CI 0.91–1.27) and LBC1936 (hazard ratio = 1.00, 95% CI 0.79–1.28). Predictors based on small sample size are prone to confounding by cellular compositions relative to those from large sample size. We observed comparable performance of our predictor in non-blood tissues with a multi-tissue-based predictor. Conclusions This study indicates that the epigenetic clock can be improved by increasing the training sample size and that its association with mortality attenuates with increased prediction of chronological age. Electronic supplementary material The online version of this article (10.1186/s13073-019-0667-1) contains supplementary material, which is available to authorized users. Keywords: DNA methylation, Age prediction, Epigenetic clock, Ageing, Mortality BACKGROUND Ageing is a major risk for diseases and mortality [1, 2]. Chronological age has been widely used as a marker of ageing due to ease and accuracy of measurement [1]. However, it is not necessarily a good predictor of biological ageing since individuals with the same chronological age can vary in health, especially in later life [3]. Therefore, researchers have attempted to search for biomarkers of ageing that can predict functional capability at a later age [4, 5]. In 2013, Hannum et al. and Horvath built chronological age predictors (termed ‘epigenetic clock’) based on DNA methylation [6, 7]. Subsequently, a number of other DNA methylation-based ‘clocks’ have been developed [8–11]. These clocks utilize age acceleration residuals (AAR, defined as the residuals from regressing predicted age on chronological age) as a biomarker of ageing[7]. Individuals with positive AAR are considered to be biologically older than their chronological age, and vice versa. A number of studies have identified associations between AAR and mortality, obesity and other health-related traits [12–15]. However, a better way of predicting these health-related traits is developing a predictor with the target trait as a reference [8, 10, 11, 16]. For example, a mitotic clock has been built for tracking the proliferation of cancer [8, 16]. DNAPhenoAge [10] and DNAmGrimAge [11] predictors were developed to predict a composite phenotype (composed of chronological age and clinical markers). Both of these predictors show stronger associations with lifespan and mortality than Horvath’s age predictor [7]. By definition, AAR from a perfect chronological age predictor should be zero. Therefore, no information on inter-individual variation in biological age can be identified based on such a predictor [17]. Nevertheless, whether we can develop a perfect chronological age predictor based on DNA methylation is unknown. Besides, whether the reported associations between AAR and health-related traits (e.g., mortality) are inflated (e.g., by potential confounders) and/or affected by the prediction accuracy of ‘epigenetic clock’ has not been investigated. In the present study, to investigate whether a perfect DNA methylation-based age predictor is theoretically available, we estimated the proportion of variance of age that could be explained by DNA methylation using a mixed linear model. We developed age predictors based on training sets with various sample sizes using Elastic Net [18] and Best Linear Unbiased Prediction (BLUP) [19]. We calculated AAR based on these age predictors and examined whether its association with mortality is affected by the prediction accuracy and potential confounders. We further investigated the performance of our predictor in samples from tissues other than blood. METHODS Study population We collected 14 data cohorts with 13,661 samples (13,402 from blood and 259 from saliva) in the age range of 2 to 104 years measured on the DNA methylation HumanMethylation450 chips and Illumina EPIC (850 K) arrays (Table 1). Eight of these cohorts were from the public domain (GEO database) and six datasets from the investigators. The six datasets include Lothian Birth Cohort (LBC) 1921/1936, Brisbane Systems Genomics Study (BSGS), Systems Genomic of Parkinson’s Disease consortium (SGPD), Motor Neuron Disease cohort (MND), and Generation Scotland (GS). Details of samples in BSGS and LBC cohorts can be found in Powell et al. [22] and Deary et al. [20, 21]. GS is a population- and family-based cohort recruited through the National Health Service (NHS) Scotland general practitioner research network [24, 25]. The SGPD cohort is from a collaborative research project on systems genomics of Parkinson’s disease. Similarly, the MND cohort is from a systems genomics study of Motor Neuron Disease in Chinese subjects (see descriptions in Benyamin et al. [23]). DNA methylation beta value at each probe was used for analysis. [Table: Description of DNA methylation cohorts] After quality control, we obtained a set of 319,607 probes (called the No Pruned set) for each sample (Additional file 1). The effective number of independent methylation probes was previously reported to be around 200 [30], indicating a dense correlation structure. Therefore, we generated a pruned probe set (128,405 probes) (Additional file 1) and compared its performance in age prediction with that based on No Pruned set. Two cohorts were identified to be outliers in the principal components analysis (PCA) using probes from the No Pruned set (Additional file 1: Figure S1). However, the prediction accuracy in both of these cohorts is not low, and thus, we kept them in the subsequent analysis (Additional file 1). Most of the training samples of our age predictors are from the blood. To test the performance of our age predictors in non-blood tissues, we downloaded 13 additional cohorts (from GEO database) with samples from tissues other than the blood (Additional file 2: Table S1). Estimating the proportion of variance of chronological age explained by DNA methylation The GS and SGPD samples were used in estimating the proportion of variance of chronological age explained by DNA methylation. Among the 5101 samples in the GS cohort, a subset of 2586 unrelated individuals, with a genetic relationship coefficient below 0.05 and with no shared nuclear family environment, were considered for the analysis. Meanwhile, we selected 1299 unrelated (genetic relationship coefficient < 0.05) individuals with available age

Read the complete text