A multiomics-clinical model to predict treatment response in psoriasis

Abstract

Psoriasis is a chronic inflammatory skin disease with highly variable responses to biologic therapies. Predictive tools to guide personalized treatment remain an unmet clinical need. To develop and validate a multiomics-clinical model for predicting a 75% reduction in the Psoriasis Area and Severity Index (PASI 75) in patients with psoriasis. In this retrospective study, 342 patients with moderate-to-severe plaque psoriasis were randomly assigned to training ( n = 240) and validation ( n = 102) sets. Candidate predictors included demographic, clinical, and multi-omics variables (e.g., HLA-Cw6 genotype, serum IL-17 A). Following univariate screening and Least Absolute Shrinkage and Selection Operator regression, a multivariable logistic regression model was constructed. Model performance was assessed using the area under the curve (AUC), calibration, and decision curve analysis, with interpretability enhanced by SHapley Additive exPlanations. Six variables were identified as independent predictors. The model achieved an AUC of 0.806 (95%CI: 0.683–0.919) in the validation set, demonstrating good calibration (Hosmer-Lemeshow, P > 0.05) and clinical net benefit. HLA-Cw6 genotype (OR = 3.165) and serum IL-17 A (OR = 1.015) were the strongest contributors. This integrated model accurately predicts PASI 75 response, providing a promising tool for personalizing treatment in psoriasis. External validation is warranted.