Machine Learning-Driven Prediction Models for Brodalumab Therapeutic Effect and Response Speed in Plaque Psoriasis

Area of Science:

• Dermatology and Computational Biology
• Genomics and Bioinformatics
• Precision Medicine

Background:

• Biologic therapies have advanced plaque psoriasis treatment.
• Individual patient responses to biologics vary significantly.
• Predictive models are needed for personalized psoriasis therapy.

Purpose of the Study:

• To develop machine learning models for predicting Brodalumab response in plaque psoriasis patients.
• To identify key genes associated with treatment response.
• To support personalized treatment strategies and resource allocation.

Main Methods:

• Transcriptomic and clinical data from 116 moderate-to-severe plaque psoriasis patients were analyzed.
• Differential gene expression and Lasso regression identified response-related genes.
• LightGBM models were trained and validated using cross-validation.

Main Results:

• Lasso identified genes in known psoriasis pathways and novel targets like WIF1 and SAA1.
• Machine learning models accurately predicted 12-week treatment response and 4-week response speed.
• Models using combined lesional and non-lesional data achieved high AUC-ROC values (up to 98.70%).

Conclusions:

• Developed models offer robust prediction of Brodalumab response.
• These tools support precision medicine in plaque psoriasis management.
• Optimized treatment selection can improve patient outcomes and resource utilization.