Varshavsky, M. | Search bioRxiv

bioRxiv2024

Time is encoded by methylation changes at clustered CpG sites

Age-dependent changes in DNA methylation allow chronological and biological age inference, but the underlying mechanisms remain unclear. Using ultra-deep sequencing of >300 blood samples from healthy individuals, we show that age-dependent DNA methylation changes are regional and occur at multiple adjacent CpG sites, either stochastically or in a coordinated block-like manner. Deep learning analysis of single-molecule patterns in two genomic loci achieved accurate age prediction with a median error of 1.46-1.7 years on held-out human blood samples, dramatically improving current epigenetic clocks. Factors such as gender, BMI, smoking and other measures of biological aging do not affect chronological age inference. Longitudinal 10-year samples revealed that early deviations from epigenetic age are maintained throughout life and subsequent changes faithfully record time. Lastly, the model inferred chronological age from as few as 50 DNA molecules, suggesting that age is encoded by individual cells. Overall, DNA methylation changes in clustered CpG sites illuminate the principles of time measurement by cells and tissues, and facilitate medical and forensic applications. O_FIG O_LINKSMALLFIG WIDTH=174 HEIGHT=200 SRC="FIGDIR/small/626674v1_ufig1.gif" ALT="Figure 1"> View larger version (46K): org.highwire.dtl.DTLVardef@c4dd52org.highwire.dtl.DTLVardef@9e6e62org.highwire.dtl.DTLVardef@160f8c3org.highwire.dtl.DTLVardef@16be7a3_HPS_FORMAT_FIGEXP M_FIG C_FIG

genomics↗

bioRxiv2023

Accurate age prediction from blood using of small set of DNA methylation sites and a cohort-based machine learning algorithm

Chronological age prediction from DNA methylation sheds light on human aging, indicates poor health and predicts lifespan. Current clocks are mostly based on linear models from hundreds of methylation sites, and are not suitable for sequencing-based data. We present GP-age, an epigenetic clock for blood, that uses a non-linear cohort-based model of 11,910 blood methylomes. Using 30 CpG sites alone, GP-age outperforms state-of-the-art models, with a median accuracy of ~2 years on held-out blood samples, for both array and sequencing-based data. We show that aging-related changes occur at multiple neighboring CpGs, with far-reaching implications on aging research at the cellular level. By training three independent clocks, we show consistent deviations between predicted and actual age, suggesting individual rates of biological aging. Overall, we provide a compact yet accurate alternative to array-based clocks for blood, with future applications in longitudinal aging research, forensic profiling, and monitoring epigenetic processes in transplantation medicine and cancer. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=158 HEIGHT=200 SRC="FIGDIR/small/524874v1_ufig1.gif" ALT="Figure 1"> View larger version (31K): org.highwire.dtl.DTLVardef@14eb626org.highwire.dtl.DTLVardef@98aecdorg.highwire.dtl.DTLVardef@1fc2ca1org.highwire.dtl.DTLVardef@d6138f_HPS_FORMAT_FIGEXP M_FIG C_FIG O_LIMachine learning analysis of a large cohort (~12K) of DNA methylomes from blood C_LIO_LIA 30-CpG regression model achieves a 2.1-year median error in predicting age C_LIO_LIImproved accuracy ([≥]1.75 years) from sequencing data, using neighboring CpGs C_LIO_LIPaves the way for easy and accurate age prediction from blood, using NGS data C_LI MotivationEpigenetic clocks that predict age from DNA methylation are a valuable tool in the research of human aging, with additional applications in forensic profiling, disease monitoring, and lifespan prediction. Most existing epigenetic clocks are based on linear models and require hundreds of methylation sites. Here, we present a compact epigenetic clock for blood, which outperforms state-of-the-art models using only 30 CpG sites. Finally, we demonstrate the applicability of our clock to sequencing-based data, with far reaching implications for a better understanding of epigenetic aging.

bioinformatics↗