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

Comprehensive Summary

The following study presents a large prospective, multicentre, multi-omic study investigating the molecular basis of psoriasis heterogeneity and disease severity. The study analysed RNA-sequencing data from skin(lesional and non-lesional) and blood samples collected from 146 patients with moderate-to-severe chronic plaque psoriasis initiating biologic therapy (adalimumab or ustekinumab), with replication in an independent cohort. Using Weighted Gene Co-expression Network Analysis (WGCNA), Independent Component Analysis (ICA), and explainable machine-learning models, the authors identified reproducible transcriptomic endotypes linked to clinical phenotyles such as BMI, HLA genotype, and PASI disease severity scores. A key finding was a 14-gene signature that was negatively associated with both BMI in non-lesional skin and disease severity in lesional skin, alongside distinct immune- and keratinocyte-driven modules that were positively or negatively associated with disease severity.

Outcomes and Implications

The study provides one of the most comprehensive molecular frameworks to date for understanding psoriasis as a heterogeneous, system-level disease rather than a single clinical entity. By integrating transcriptomics with explainable Gaussian process regression, the authors demonstrate that disease severity can be accurately predicted using compact gene signatures, outperforming models based solely on clinical variables. Importantly, the work reveals how genetic factors and environmental modifiers such as obesity shape inflammatory pathways across tissue compartments. Clinically, these findings move the field closer to precision dermatology, enabling patient stratification by molecular endotype, improving prediction of disease severity, and potentially enabling more tailored selection of biologic therapies. The study also highlights the value of interpretable AI models in translational medicine and supports their future integration into clinical decision support systems for chronic inflammatory skin diseases.