Fig. 2: Systems-level gene modules and latent factors correlate with disease and disease severity endotypes in psoriasis. | Communications Medicine

Fig. 2: Systems-level gene modules and latent factors correlate with disease and disease severity endotypes in psoriasis.

From: Transcriptomic profiling and machine learning uncover gene signatures of psoriasis endotypes and disease severity

Fig. 2: Systems-level gene modules and latent factors correlate with disease and disease severity endotypes in psoriasis.

Heatmaps display pearson correlations between (upper panels) WGCNA module eigengenes or (lower panels) ICA latent factors and clinical/demographic variables for (a) skin and (b) blood in the discovery cohort. Benjamini–Hochberg multiple testing correction was done separately for modules and factors and skin and blood. Black outlines denote correlations significant at false-discovery rate (FDR) ≤ 0.05; asterisks indicate the correlation replicates (same sign, P < 0.05) in the independent replication cohort. Only modules/factors with at least one replicated correlation are shown; descriptors summarise the dominant biological theme of each module/factor. Panels (c–e) illustrate three examples: Baseline clinical endotypes—Pearson correlations with BMI in skin and HLA-C*06:02 genotype in blood (c). Disease severity endotypes—non-linear relationships between eigengenes and PASI across all visits in blood (d) and skin (e). Significant disease severity associations were first identified by pearson correlation and then visualised with spline curves. Points are coloured by visit (blue = week 0, green = week 1, red = week 4, orange = week 12). Violin plots include a median line. Curves represent natural-spline fits (3 d.f.) with 95% confidence bands, as in Fig. 1 plus. HLA human-leucocyte antigen, FDR false-discovery rate.

Back to article page