Objective To identify potential therapeutic targets for idiopathic pulmonary fibrosis (IPF) and predict related herbal medicines by integrating bioinformatics, Mendelian randomization (MR), and machine learning approaches. Methods IPF microarray datasets were obtained from the GEO database to identify differentially expressed genes (DEGs). Using expression quantitative trait loci (eQTL) data and genome-wide association study (GWAS) data, MR analysis was conducted to screen for genes associated with IPF. The risk genes identified from MR analysis were intersected with DEGs to filter core IPF-related genes. Subsequent evaluations included functional enrichment analysis, gene set enrichment analysis (GSEA), immune cell infiltration analysis, and single-cell RNA sequencing. Machine learning algorithms were applied to select optimal diagnostic feature genes. An independent GEO cohort was used for differential expression validation and receiver operating characteristic (ROC) analysis. Results A total of 916 IPF-associated DEGs were identified. Intersection with 224 MR risk genes yielded seven key genes: IRF7, TTC32, IFI6, and ISG15 (risk genes), along with ZNF204P, ISOC1, and CTSK (protective genes). These genes were primarily enriched in pathways related to interferon-beta production, RIG-I-like receptor signaling, Toll-like receptor signaling, and type I interferon signaling. Immune infiltration analysis revealed a decrease in M1/M0 macrophages and resting mast cells, alongside an increase in activated mast cells in IPF tissues. Single-cell RNA sequencing demonstrated specific expression patterns of these genes within epithelial cell subpopulations. Machine learning algorithms identified ZNF204P and IRF7 as the optimal diagnostic genes. Validation in the dataset confirmed their significant differential expression in IPF and high diagnostic accuracy. A total of 385 traditional Chinese medicines (TCMs) related to the key genes were predicted. The primary properties of these TCMs were cold, warm, and neutral (four natures); their main flavors were bitter, sweet, and pungent (five flavors); the principal meridian tropisms were the liver, lung, stomach, spleen, and kidney meridians; and the major classifications were heat-clearing drugs, tonifying drugs, blood-activating and stasis-resolving drugs, exterior-releasing drugs, and dampness-draining diuretics. Molecular docking simulations revealed that these chemical components could form stable interactions with the core proteins. Conclusion This study identified seven key genes associated with IPF. Machine learning screened ZNF204P and IRF7 as robust diagnostic biomarkers with therapeutic target potential. Furthermore, TCMs such as Zhizi (Gardeniae Fructus), Chaihu (Bupleuri Radix), Roucongrong (Cistanches Herba), and Huangqi (Astragali Radix) might be potential TCMs that targets core genes associated with IPF.

bioinformatics  /  Mendelian randomization  /  expression quantitative trait loci  /  machine learning  /  idiopathic pulmonary fibrosis  /  traditional Chinese medicine prediction