Fig. 2
Characterization of DeGAs latent structures. a, b Components from truncated singular value decomposition (TSVD) corresponds to principal components in the phenotype (a) and variant (b) spaces. The first two components of all the variants, excluding the MHC region, and relevant phenotypes are shown. b For variant PCA, we show biplot arrows (red) for selected phenotypes to help interpretation of the direction of principal components (Methods). The variants are labeled based on the genomic positions and the corresponding gene symbols. For example, "16:53813367 (FTO)" indicates the variant in gene FTO at position 53813367 on chromosome 16. c, d Phenotype (c) and gene (d) contribution scores for the first five components. PC1 is driven by largest part of the body mass that accounts for the “healthy part” (main text) including whole-body fat-free mass and genetic variants on FTO and DLEU1, whereas PC2 is driven by fat-related measurements, PC3 is driven by bioelectrical impedance measurements, PC4 is driven by eye measurements, and PC5 is driven by bioelectrical impedance and spirometry measurements along with the corresponding genetic variants (main text, Supplementary Table 2, Supplementary Data 2). Each colored segment represents a phenotype or gene with at least 0.5% and 0.05% of phenotype and gene contribution scores, respectively, and the rest is aggregated as others on the top of the stacked bar plots. The major contributing phenotype groups (Methods, Supplementary Table 2) and additional top 10 phenotypes and the top 10 genes for each component are annotated in c and d, respectively. pred.: predicted, %: percentage, mass/% mass and percentage, BP: blood pressure, AR: automated reading, L: left, R: right. Source data are provided as a Source Data file (a, b) and in Supplementary Data 2 (c, d)
