Abstract / Summary
Background: Patients with schizophrenia (SCZ) receiving antipsychotic treatment frequently develop metabolic complications, yet the mechanisms underlying liver-immune crosstalk remain poorly characterized. Methods: This study recruited 340 antipsychotic-treated SCZ patients and 340 matched healthy controls, with a 2-year follow-up restricted to the patient cohort. Candidate genes with hepatic metabolic relevance were screened from the independent GSE21138 brain transcriptomic dataset using machine learning. Molecular docking was performed to predict binding interactions of quercetin as a candidate bioactive compound. Results: Long-term antipsychotic exposure was associated with elevated serum 9-hydroxyrisperidone (9-OH-RISP) and substantial changes in liver function and metabolic markers; drug metabolite concentrations correlated with hepatic indices. Transcriptomic analysis identified 317 differentially expressed genes, from which 26 genes with liver-metabolism-related functions were selected as candidates. Seven hub genes (ABHD6, ALDH4A1, ETNPPL, etc.) were further identified via Support Vector Machine and Random Forest algorithms, and these genes were enriched in metabolic pathways, ferroptosis and other key signaling cascades. Unsupervised clustering based on hub gene expression stratified patients into high-inflammation (C2) and low-inflammation (C1) subtypes. The C2 subtype displayed computationally estimated enrichment of immune cells including B cells and naive CD8+ T cells. Molecular docking predicted strong binding affinity (binding energy ≤ −7.0 kcal/mol) between quercetin and proteins encoded by six hub genes, implying potential multi-target anti-inflammatory and metabolic–regulatory interactions. Conclusions: This study highlights the involvement of liver–immune crosstalk in antipsychotic-associated metabolic perturbations. The findings suggest exploratory potential for quercetin as a candidate bioactive compound in high-inflammatory SCZ subtypes, though these in silico predictions require further experimental validation.