Clinical AI-assisted FXR prioritization and WADDAICA-guided design of a berberine-inspired analogue for MASH: integrated computational and experimental validation.

BACKGROUND: Metabolic dysfunction-associated steatotic liver disease (MASLD) and metabolic dysfunction-associated steatohepatitis (MASH), lack effective targeted therapies despite increasing global prevalence. The farnesoid X receptor (FXR) is a key regulator of hepatic lipid metabolism, inflammation, and fibrosis, making it a promising therapeutic target. However, existing FXR modulators are associated with limited efficacy and adverse effects. OBJECTIVE: This study aimed to establish a sequential clinical-to-molecular framework combining clinical AI-assisted liver-disease feature learning, FXR target prioritization, WADDAICA-guided molecular design, and experimental validation of a berberine-inspired analogue. METHODS: An SE attention-based Transformer model was trained on the Indian Liver Patient Dataset only for liver-disease classification and clinical feature-prioritization analysis. The model output consisted of disease-classification probability and feature-importance information; it was not used for compound optimization. Molecular design of the berberine-inspired analogue was performed independently using the WADDAICA platform. RESULTS: The AI model demonstrated excellent performance (AUC = 0.9998), identifying clinically relevant metabolic markers linked to FXR pathways. The WADDAICA-designed berberine-inspired analogue showed favorable predicted interactions with FXR, supported by MD and DFT analyses. Favorable pharmacokinetic properties, including high gastrointestinal absorption and low blood-brain barrier permeability, were observed. In vitro, the compound significantly improved cell viability, reduced lipid accumulation, and modulated key molecular markers, including upregulation of FXR and PPARα and downregulation of SREBP-1c, TNF-α, IL-6, and fibrotic markers. CONCLUSION: This study presents a multi-scale, AI-integrated pipeline linking clinical prediction with molecular design and biological validation. The WADDAICA-designed berberine-inspired analogue showed FXR-associated molecular changes and preliminary cytoprotective and anti-steatotic effects in an FFA-induced HepG2 model. These findings support further investigation of its putative FXR-mediated mechanism of action using dedicated receptor functional assays.