Machine Learning Algorithms as Tools for Identifying Predictive Autoantibody Biomarkers in Pemphigus Vulgaris

This study investigates the role of non-Dsg autoantibodies in Pemphigus vulgaris (PV) using protein microarray technology on 633 serum samples (421 patients, 212 controls). The analysis identified increased reactivity to 37 autoantigens in active PV patients compared to controls, with significant differences based on HLA haplotypes. K-Nearest Neighbor (KNN) effectively distinguished between different patient groups, and longitudinal analysis showed unique autoantigen profiles for each patient. The findings suggest that both Dsg and non-Dsg autoantibodies play a role in PV pathogenesis and highlight the influence of HLA genetics, offering insights into disease mechanisms and potential predictive biomarkers for personalized therapies.