INTRODUCTIONWhile the global burden of inflammatory bowel diseases (IBD) is increasing, the identification of novel therapeutic targets and biomarkers is of significant importance. In particular, blood transcriptomes provide a non-invasive source for biomarker discovery. Therefore, this study aimed to identify potential blood markers for IBD.METHODSBy employing an integrated transcriptomics approach, four datasets obtained from blood specimens of patients with IBD were analyzed (GSE119600, GSE94648, GSE86434, and GSE71730). After determining differentially expressed genes (DEGs) in IBD, a protein-protein interaction (PPI) network was constructed, and regulatory miRNAs targeting hub genes were identified. Weighted gene co-expression network analysis (WGCNA) was carried out to determine IBD-specific modules. Subsequently, converging results from differential expression analysis and WGCNA were subjected to random forest (RF) decision tree-based and LASSO regression methods. Lastly, the diagnostic efficacy of genes highlighted by both machine learning methods was measured using receiver operating characteristic (ROC) analysis in the integrated dataset, in each individual dataset separately, and in external datasets (GSE276395, GSE169568, GSE112057, GSE100833, GSE33943, and GSE3365).RESULTSDownregulated TNF was identified as the central hub gene of the PPI network, and PRF1 was the only gene identified as a hub gene in a co-expressed gene module enriched in IBD. Following the identification of FEZ1 and NLRC5 among the top 10 genes by both RF and LASSO, ROC analysis demonstrated their acceptable diagnostic efficacy in the integrated data. However, only FEZ1 was considered a potential biomarker based on replication of the results in the external datasets.CONCLUSIONSThe results of the present study suggest FEZ1 as a potential blood biomarker for IBD. While autophagy is currently the most convincing explanation for the involvement of FEZ1 in IBD, further investigations are required to elucidate its immunological role.