Immune Microenvironment Characterization and Machine Learning-Guided Identification of Diagnostic Biomarkers for Ulcerative Colitis.

Background: Ulcerative colitis (UC) is a chronic inflammatory bowel disease hallmarked by dysregulated immune responses. Current treatments often show limited efficacy, highlighting the need for novel diagnostic and therapeutic approaches.

Methods: RNA-Seq data from 495 UC patients and 320 controls (training dataset) and 389 UC patients and 209 controls (testing dataset) were analyzed. Immune cell infiltration was assessed via the ImmuCellAI algorithm, while differential expression analysis and WGCNA were performed to identify key immune-related genes. Moreover, machine learning models, including Random Forest and Best Subset Selection, were used to construct and validate an optimal diagnostic framework. Lastly, the findings were further corroborated using immunohistochemistry conducted on tissue samples from UC patients and controls.

Results: Thirteen immune cell types, including B cells, macrophages, and naive CD4+ T cells, were identified as significantly altered in UC. Likewise, cytokines such as IL-10, TGF-β, RORγ, and IL-21 exhibited abnormal expression patterns in UC tissues. WGCNA identified three immune cell-associated gene modules, among which the MEblue, MEturquoise, and MEgrey modules were highly correlated with aberrant immune cells. Additionally, machine learning models identified 99 candidate genes, from which an optimal diagnostic model comprising eight crucial genes (GATA2, IL8, LAT, NOLC1, SMARCA5, SMC3, STX10, ZMIZ1) was constructed, achieving an AUC of 0.964 in the training dataset, 0.926 in the internal test dataset, and 0.884 in the independent test dataset. Functional enrichment analysis revealed associations with inflammatory and immune-regulatory pathways, highlighting their biological relevance. Moreover, the identified eight genes hold translational potential for clinical diagnostics and may serve as a foundation for future precision-targeted therapies in UC.

Conclusion: This study highlights alterations in the immune microenvironment in UC and presents an accurate eight-gene diagnostic model, offering the potential for early detection and novel therapeutic targets.