Combined Salivary Proteome Profiling and Machine Learning Analysis Provides Insight into Molecular Signature for Autoimmune Liver Diseases Classification.

Abstract

Autoimmune hepatitis (AIH) and primary biliary cholangitis (PBC) are autoimmune liver diseases that target the liver and have a wide spectrum of presentation. A global overview of quantitative variations on the salivary proteome in presence of these two pathologies is investigated in this study. The acid-insoluble salivary fraction of AIH and PBC patients, and healthy controls (HCs), was analyzed using a gel-based bottom-up proteomic approach combined with a robust machine learning statistical analysis of the dataset. The abundance of Arginase, Junction plakoglobin, Desmoplakin, Hexokinase-3 and Desmocollin-1 decreased, while that of BPI fold-containing family A member 2 increased in AIHp compared to HCs; the abundance of Gelsolin, CD14, Tumor-associated calcium signal transducer 2, Clusterin, Heterogeneous nuclear ribonucleoproteins A2/B1, Cofilin-1 and BPI fold-containing family B member 2 increased in PBCp compared to HCs. The abundance of Hornerin decreased in both AIHp and PBCp with respect to HCs and provided an area under the ROC curve of 0.939. Machine learning analysis confirmed the feasibility of the salivary proteome to discriminate groups of subjects based on AIH or PBC occurrence as previously suggested by our group. The topology-based functional enrichment analysis performed on these potential salivary biomarkers highlights an enrichment of terms mostly related to the immune system, but also with a strong involvement in liver fibrosis process and with antimicrobial activity.