Comparative Analysis of Machine Learning Approaches for Molecular Pathway Identification and Biomarker Discovery in Immune-Related Diseases

Abstract

Immune-related diseases pose significant diagnostic challenges due to complex molecular mechanisms and heterogeneous clinical presentations. Machine learning approaches have emerged as powerful tools for molecular pathway identification and biomarker discovery. This comparative study evaluates five machine learning algorithms using transcriptomic datasets from rheumatoid arthritis, systemic lupus erythematosus, and inflammatory bowel disease. We assess algorithm performance across accuracy, computational efficiency, biological relevance, and clinical validity. Graph neural networks achieved superior disease classification performance (AUC: 0.847) and identified 21 significantly enriched pathways, compared to traditional clustering methods (classification AUC: 0.762, 17 pathways identified). Results establish practical guidelines for algorithm selection, advancing personalized diagnostic development.