Mus Musculus) can be slowed by gold standard anti-aging interventions such as calorie restriction and growth hormone receptor knock-outs. Using DNA methylation data from previous publications with data collected in house for a total 1189 samples spanning 193,651 CpG sites, we developed 4 novel epigenetic clocks by choosing different regression models (elastic net- versus ridge regression) and by considering different sets of CpGs (all CpGs vs highly conserved CpGs). We demonstrate that accurate age estimators can be built on the basis of highly conserved CpGs. However, the most accurate clock results from applying elastic net regression to all CpGs. While the anti-aging effect of calorie restriction could be detected with all types of epigenetic clocks, only ridge regression based clocks replicated the finding of slow epigenetic aging effects in dwarf mice. Overall, this study demonstrates that there are trade-offs when it comes to epigenetic clocks in mice. Highly accurate clocks might not be optimal for detecting the beneficial effects of anti-aging interventions." name="description"> A multi-tissue full lifespan epigenetic clock for mice - Figure f1 | Aging
← Back
Research Paper Volume 10, Issue 10 pp 2832—2854

A multi-tissue full lifespan epigenetic clock for mice

class="figure-viewer-img"

Figure 1. Accuracy of ridge regression epigenetic age predictions. DNA methylation age (y-axis) versus chronological age (x-axis) for all mouse samples. (a) Performance of ridge regression clock based on all 192K CpGs in all training samples. The training set estimates of the accuracy are overly optimistic and should be ignored. (b) Results by tissue type of cross-validated predictions obtained by iteratively withholding one “batch” (tissue x publication). For the batch cross-validation of this clock, the global Pearson correlation between predicted and chronological age was 0.79 (p < 2E-195) with a mae of 3.1 months. All models in these iterative cross-validations had the same size of 193,651 CpGs. (c) Scatter plots by tissue type based on DNAm age estimates made with an iterative leave-one-sample-out cross-validation. The correlation between predicted and chronologic age was 0.85 (p < 6E-258) with a mae of 2.1 months.