An LSEC-focused computational drug repurposing platform for liver fibrosis: Identification of vorinostat and other LSEC-protective candidates.

Liver sinusoidal endothelial cells (LSECs) are increasingly recognized as a critical yet underexplored cell type in anti-fibrotic drug development. This study presents a computational drug screening platform integrating LSEC-specific transcriptomic analysis across simple steatosis, fibrotic nonalcoholic steatohepatitis (NASH), and cirrhosis, with tiered gene signature selection combining machine learning, large language model-assisted curation, gene safety assessment, and Connectivity Map-based screening using human endothelial perturbational profiles. The platform identifies 6 clinical-stage and 8 preclinical candidates with LSEC-protective potential. Among these, vorinostat (SAHA), a clinically approved histone deacetylase (HDAC) inhibitor, is selected for experimental validation. In hepatocyte-specific Asah1-deficient mice fed a Paigen diet, SAHA attenuates hepatic inflammation, fibrosis, LSEC dysfunction, and portal hemodynamic abnormalities, with effects confirmed in a hepatotoxin (CCl4)-induced fibrosis model. High mobility group box 1 (HMGB1) is identified as a key hepatocyte-derived paracrine mediator of LSEC injury through Transwell co-culture and glycyrrhizin rescue. Vorinostat dose-dependently reverses HMGB1-induced LSEC dysfunction across inflammation, capillarization, fibrogenesis, and vasoconstriction, associated with endothelial transcription factor reprogramming including KLF2 upregulation, validated in primary LSECs and in vivo. SAHA also protected LSECs from TNF-α-induced inflammation and reduced monocyte adhesion. These findings establish an LSEC-focused drug repurposing framework and identify candidates for LSEC-protective anti-fibrotic therapy.