Acute alcoholic liver injury is closely associated with oxidative stress and inflammation. Liuwei Wuling tablet (LWWL), a traditional Chinese medicine formula, has been clinically used for liver injury treatment; however, its multicomponent and multitarget characteristics make the underlying mechanisms difficult to elucidate using conventional approaches. Network pharmacology combined with machine learning provides a systematic strategy to predict core targets of complex herbal formulas, offering a basis for subsequent experimental validation.

Active components of LWWL were screened from TCMSP and BATMAN-TCM databases. Known targets of these components were retrieved from TCMSP and BATMAN-TCM, while additional targets were predicted using PharmMapper; all targets were integrated as the potential target set of LWWL. Disease-related targets for acute alcoholic liver injury were obtained from GeneCards, OMIM, and DisGeNET databases. Intersection targets between LWWL and disease were identified as potential therapeutic targets. A protein-protein interaction network was constructed to screen core targets, followed by KEGG pathway enrichment analysis. Random forest algorithm was employed to calculate importance scores for each target, and key targets associated with oxidative stress and inflammatory pathways were selected based on importance ranking combined with enrichment results. An acute alcoholic liver injury mouse model was established, and LWWL was administered. Serum ALT/AST levels, hepatic SOD/GSH activities, MDA/TG contents, and TNF-α/IL-1β levels were measured. Hepatic pathological changes were observed by HE and Oil Red O staining.

A total of 59 active components were identified, corresponding to 347 targets. Disease-related targets from three databases yielded 518 targets, with 89 overlapping targets between LWWL and disease. PPI network analysis revealed STAT3, AKT1, and MAPK1 as core targets, which were primarily enriched in PI3K-AKT, TNF, MAPK, FoxO, and AMPK signaling pathways. Random forest importance scoring identified 10 key targets including STAT3, TNF, and MAPK1. Animal experiments demonstrated that LWWL significantly decreased serum ALT/AST levels, increased hepatic SOD/GSH activities, reduced MDA/TG contents, and downregulated TNF-α/IL-1β levels, along with ameliorated hepatocyte injury, consistent with the predictive findings.

This study employed network pharmacology combined with random forest algorithm to predict that LWWL may exert antioxidant and anti-inflammatory effects against acute alcoholic liver injury through regulation of PI3K-AKT, TNF, FoxO/AMPK pathways, which were subsequently validated by animal experiments. The "prediction-screening-validation" strategy provides a methodological reference for mechanism studies of traditional Chinese medicine formulas.