Machine learning-based algorithm identifies key mitochondria-related genes in non-alcoholic steatohepatitis.

BACKGROUND: Evidence suggests that hepatocyte mitochondrial dysfunction leads to abnormal lipid metabolism, redox imbalance, and programmed cell death, driving the onset and progression of non-alcoholic steatohepatitis (NASH). Identifying hub mitochondrial genes linked to NASH may unveil potential therapeutic targets. METHODS: Mitochondrial hub genes implicated in NASH were identified via analysis using 134 algorithms. RESULTS: The Random Forest algorithm (RF), the most effective among the 134 algorithms, identified three genes: Aldo-keto reductase family 1 member B10 (AKR1B10), thymidylate synthase (TYMS), and triggering receptor expressed in myeloid cell 2 (TREM2). They were upregulated and positively associated with genes promoting inflammation, genes involved in lipid synthesis, fibrosis, and nonalcoholic steatohepatitis activity scores in patients with NASH. Moreover, using these three genes, patients with NASH were accurately categorized into cluster 1, exhibiting heightened disease severity, and cluster 2, distinguished by milder disease activity. CONCLUSION: These three genes are pivotal mitochondrial genes implicated in NASH progression.

Machine learning-based integration identifies the ferroptosis hub genes in nonalcoholic steatohepatitis.

BACKGROUND: Ferroptosis, is characterized by lipid peroxidation of fatty acids in the presence of iron ions, which leads to cell apoptosis. This leads to the disruption of metabolic pathways, ultimately resulting in liver dysfunction. Although ferroptosis is linked to nonalcoholic steatohepatitis (NASH), understanding the key ferroptosis-related genes (FRGs) involved in NASH remains incomplete. NASH may be targeted therapeutically by identifying the genes responsible for ferroptosis. METHODS: To identify ferroptosis-related genes and develop a ferroptosis-related signature (FeRS), 113 machine-learning algorithm combinations were used. RESULTS: The FeRS constructed using the Generalized Linear Model Boosting algorithm and Gradient Boosting Machine algorithms exhibited the best prediction performance for NASH. Eight FRGs, with ZFP36 identified by the algorithms as the most crucial, were incorporated into in FeRS. ZFP36 is significantly enriched in various immune cell types and exhibits significant positive correlations with most immune signatures. CONCLUSION: ZFP36 is a key FRG involved in NASH pathogenesis.