Fig. 4: Gene-SGAN identifies five subtypes of brain changes related to hypertension (H1-H5).

a The five subtypes show distinct imaging patterns. Voxel-wise group comparisons (two-sided t-test) were performed between HC participants (i.e., non-hypertensive participants) and participants assigned to H1, H2, H3, H4, and H5, respectively. False discovery rate (FDR) correction for multiple comparisons with a p-value threshold of 0.05 was applied. Warmer color denotes brain atrophy (i.e., HC > subtype), and cooler color represents larger tissue volume (i.e., subtype > HC). b The five subtypes show distinct clinical, cognitive, and demographic characteristics. Box and whisker plots and bar plots show the characteristic of demographic, clinical, and cognitive variables. In TMT B plots, outliers are excluded for visualization purposes. (center line, median; red marker: mean; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; points, outliers). Sample sizes of each variable are presented below each figure. c The five subtypes show different genetic architectures. The Manhattan plot displays significant SNP-subtype associations among 117 hypertension-associated SNPs (one tailed likelihood-ratio test with multinomial logistic regression models) using all hypertensive patients’ data. The two dashed lines denote the p-value thresholds of 0.05 after adjusting for multiple comparisons via Bonferroni (top) and B-H methods (bottom). We manually annotated significant SNPs that survived the Bonferroni correction with the SNP numbers and the mapped genes via their physical positions. In this figure, we defined the effective allele of each SNP as the allele positively associated with hypertension or WMH (SNPs identified in hypertension-gene interaction analyses⁵⁴) reported in previous literature. The bar plots reveal EAFs of the top nine significant SNPs within each subtype.