Identification of key genes as diagnostic biomarkers for IBD using bioinformatics and machine learning

作者：Tianhao Li, Haoren Jing, Xinyu Gao, T. Zhang, Haitao Yao, Xipeng Zhang, Mingqing Zhang · 发表于：Journal of Translational Medicine · 年份：2025 · DOI：10.1186/s12967-025-06531-1 · 被引用次数：24 · 研究领域：Inflammatory Bowel Disease、Genetic Associations and Epidemiology、Ferroptosis and cancer prognosis

BACKGROUND: The pathogenesis of inflammatory bowel disease (IBD) involves complex molecular mechanisms, and achieving clinical remission remains challenging. This study aims to identify IBD-potential biomarkers, analyze their correlation with immune cell infiltration, and identify genes that have a causal relationship with IBD. METHODS: RNA-seq datasets for IBD were retrieved from GEO, stratified into discovery (GSE75214), validation (GSE36807), and independent testing cohorts (GSE179285, GSE47908). Through comparative expression profiling of the discovery cohort, IBD-associated differentially expressed genes (DEGs) were detected. Core candidate genes were subsequently prioritized using protein-protein interaction network analysis, further refined through machine learning approaches (Random Forest/Support Vector Machines). Immune cell abundance quantification and statistical correlation analyses with IBD-associated transcripts were conducted via the CIBERSORTx deconvolution algorithm. To complement these findings, blood expression quantitative trait loci (eQTL) data from GTExv8.ALL.Whole_Blood were integrated with IBD genome-wide association statistics from the FinnGen consortium. This multi-omics integration framework employed: (1) Bayesian colocalization to assess shared causal variants, (2) HEIDI heterogeneity testing, and (3) summary Mendelian randomization (SMR) for causal inference validation. RESULTS: Three genes, IRF1, GBP5, and PARP9, demonstrated significant IBD-pro...