Bioinformatics-based identification of gut microbiota-related candidate genes and computational drug prediction for peripheral artery disease

Peripheral artery disease (PAD) is a highly prevalent cardiovascular disorder that threatens global public health. Accumulating clinical and omics evidence suggests potential crosstalk between the gut microbiota and PAD progression. However, the host genes that may mediate these associations, their cell-type-specific regulatory roles, and corresponding therapeutic candidates remain poorly characterized. Gut microbiota-associated genes and PAD-related genes were separately retrieved from the gutMGene and GeneCards databases. Overlapping genes were evaluated using multiple machine-learning algorithms combined with SHapley Additive exPlanations (SHAP) to identify core candidate genes. Single-cell RNA-seq datasets were analyzed to profile cell-type-specific gene expression patterns, and CellChat was used to infer ligand-receptor-mediated cell-cell communication. CIBERSORT was used to estimate immune-cell fractions. A transcriptome-based risk nomogram based on core genes was developed for PAD risk prediction. Candidate compounds targeting core proteins were screened from the DSigDB database and further assessed by molecular docking. Finally, RT-qPCR was performed on muscle tissues from a hindlimb ischemia mouse model to verify differential expression of core genes. Eighty-nine overlapping genes linking gut microbiota signatures to PAD were identified and enriched for inflammation, lipid metabolism, and atherosclerosis-related pathways. Four core genes (GOT1, ITCH, IL1B, JUP) were prioritized. Single-cell analyses revealed cell-type-specific expression and predicted remodeling of intercellular communication networks, with core-gene expression correlated with altered immune-cell infiltration. The established nomogram showed favorable predictive performance in independent cohorts (AUC 0.751–0.968). Five candidate compounds, including curcumin, EPA, and EGCG, were prioritized via computational docking. Core-gene differential expression was further validated in vivo. This study integrated multi-omics data and machine-learning strategies to prioritize host core genes associated with the gut microbiota and PAD, established a transcriptome-derived nomogram for PAD risk stratification, and screened potential small-molecule agents using molecular docking. Our findings provide testable molecular targets and computational hypotheses for future mechanistic research and individualized PAD risk assessment.