Figure 1.
MSA design, trait representation, and benchmarking
(A) Overview of methylation technologies across genome coverage, cost, and throughput.
(B) MSA design schematic illustrating the design process. From the designable probe pool (left), CpGs whose methylations are associated with diverse methylation biology and human traits were identified (right).
(C) Major trait categories (red) and representative sub-traits (yellow) included in MSA; some traits may appear multiple times due to cohort differences.
(D) Top: MSA and EPICv2 probe enrichment EWAS hits ranked by the number of trait associations. Bottom: heatmap showing the enrichment (log2 odds ratio) of major trait group probes on MSA vs. EPICv2 and random Infinium probes.
(E) Number of CpGs per cell-type contrast on MSA vs. EPICv2 for contrasts with <500 high-quality whole-genome markers.
(F) Gene Ontology (GO) term enrichment (hypergeometric test) for genes linked to CpH probes (minimum two probes per gene) on MSA and EPICv2.
(G) Heatmap of beta value correlations between cell lines profiled by MSA. “Sample source” indicates the culturing lab.
(H) Density plots of measured beta values for methylation titration standards.
(I) Heatmap of beta value correlations between MSA (columns) with EM-seq (row) profiles for the same cell line samples. “Sample source” indicates culturing lab.
(J) Tissue prediction scores using an EPIC prediction model on MSA tissue profiles (columns). Missing EPIC probes were substituted with MSA nearest-neighbor probes.