Integrated Bioinformatics Identification and Multi-Dimensional Characterization of Diagnostic Hub Genes in Colorectal Cancer

Abstract

Colorectal cancer (CRC) motivates the investigation of complementary molecular markers, although tissue-transcriptomic discrimination does not by itself establish early-detection performance. We identified differentially expressed genes in 98 paired CRC and adjacent-normal samples from GSE44076 using gene-level limma analysis (|log2FC| > 1; Benjamini–Hochberg-adjusted p < 0.05), screened candidates through a STRING protein–protein interaction network, and evaluated diagnostic performance with independent validation in GSE9348. Functional characterization included MSigDB KEGG_LEGACY over-representation analysis and preranked GSEA, immune-marker correlations, mRNA stemness index analysis, exploratory topology-filtered cysteine assessment, exploratory TCGA-COAD prognostic modelling, and literature-curated single-cell contextualization. Seven candidates—AQP8, CA7, GUCA2A, GUCA2B, BEST4, TMIGD1, and OTOP2—showed AUC values of 0.986–0.995 in the discovery cohort; five genes showed AUC values of 0.999–1.000 in the retrospective GSE9348 tissue cohort. AQP8 and GUCA2B showed weak nominal inverse correlations with mRNAsi that did not remain significant after correction across six tests (BH-FDR = 0.078), and the genes were predominantly associated with differentiated colonic epithelial lineages. The exploratory prognostic model was non-significant. These results identify seven tissue-transcriptomic candidates with strong dataset-specific discriminative performance and provide a multidimensional biological framework warranting experimental validation and prospective clinical evaluation.