Machine learning-based prediction model for drug target identification and MASH improvement: a comprehensive analysis of biochemical and ferroptosis/autophagy biomarkers.

Metabolic dysfunction-associated steatotic liver disease (MASLD), marked by excess fat in the liver, has become the most prevalent chronic liver disease worldwide, affecting over 30% of adults. Its advanced form, metabolic dysfunction associated steatohepatitis (MASH), includes liver ballooning, and inflammation, and can progress to cirrhosis and hepatocellular carcinoma (HCC). Despite the increasing burden, effective pharmacotherapies for MASLD/MASH are still lacking. Programmed cell death mechanisms, such as autophagy and ferroptosis, are critical in the pathology of MASLD, influencing liver inflammation, fibrosis, and malignant transformation. This study employed six machine learning models—Random Forest, Logistic Regression, Extra Trees Classifier, Linear Discriminant Analysis, and Light Gradient Boosting Machine—to identify significant drug targets using Febuxostat, Perindopril, Amlodipine, and Atorvastatin, evaluated through molecular, biochemical, immunohistochemical, and pathological markers. We identified genes associated with MASH using microarray datasets from the Gene Expression Omnibus database, followed by protein-protein interaction and functional enrichment analyses to select genes related to ferroptosis, autophagy, and their epigenetic regulators (miRNAs-LncRNAs) in MASH-induced rats. Quantitative real-time PCR validated the expression of selected networks (mRNAs-miRNAs-LncRNAs). Additionally, we measured biochemical, inflammatory, and liver pathology markers to ensure the model’s robustness. Our results identified 16 out of 29 valuable therapeutic targets with an accuracy of 88.74% and an AUC of 0.9745, including LPCAT3, HGS, TSG101, SNF8, rno-miR-27a-5p, rno-miR-329-5p, CTBP1-AS2, ALT, AST, ALP, GGT, D. Bilirubin, Albumin, TMAO, GPX4, and TGFβ1.