Fig. 4: Illustration of the filtration process for DNAm-based multi-omic features used in the development of OMICmAge.
From: OMICmAge quantifies biological age by integrating multi-omics with electronic medical records
a, Diagram describing the filtration and selection of EBPs included in the OMICmAge model. b, Histograms show the full distributions of feature-wise Pearson correlations to EMRAge for three data modalities: 2,098 proteins, 1,459 metabolites (collapsed to 286 clusters via hierarchical clustering) and 46 clinical variables. Gray bars represent the unfiltered ‘background’ for each modality. Colored overlays (green indicates protein EBP, orange indicates metabolite EBP, and black indicates clinical EBP) show the subset of features meeting our a priori filter ( | ρ | > 0.1 and P value < 0.05), yielding 421 total features (protein, n = 110; metabolite, n = 286; clinical, n = 25). c, Of the 421 EMRAge-correlated candidates, we next required that predicted epigenetic biomarkers (EBPs) also correlate with their measured counterparts at Pearson ρ > 0.2 (P value < 0.05). Histograms again show the EMRAge-selected background distributions in gray, with colored overlays (green, protein; orange, metabolite; black, clinical) indicating the features that passed this second filter (protein, n = 109; metabolite, n = 266; clinical, n = 21). Pearson correlation coefficients for each multi-omic feature and EBP are reported in Supplementary Tables 10 and 11.
