Abstract
BACKGROUND: Colorectal cancer (CRC) is one of the leading contributors to cancer-related deaths worldwide, with more than 900,000 new diagnoses and related deaths each year. This study aims to explore the prognostic value of tumor-infiltrating immune cell (TIIC)-related genes in CRC, in order to discover new biomarkers and therapeutic targets. METHODS: We integrated CRC transcriptome data from public databases to construct and validate a prognostic model and analyzed single-cell RNA sequencing (scRNA-seq) data to classify immune cell subtypes. A suite of computational models was employed to assess TIIC signature scores and to refine the selection of prognostic TIIC-related genes using multiple machine learning techniques-including Random Survival Forest (RSF), LASSO regression, and Cox proportional hazards regression, among others. In addition, pathway enrichment, immune signature difference analyses, and immunotherapy response predictions were performed. Potential biomarkers and therapeutic targets were identified through differential gene analysis, gene set enrichment analysis (GSEA), and copy number variation (CNV) landscape comparisons between high and low TIIC groups. RESULTS: We identified 137 significant TIIC-RNAs within the CRC microenvironment and developed a prognostic model based on five key TIIC-RNAs. This model, which leveraged machine learning methods such as RSF, LASSO, and Cox regression, demonstrated outstanding performance in survival prediction across TCGA-CRC and external validation datasets, outperforming 22 existing prognostic models. Furthermore, the high TIIC score group showed heightened expression of angiogenesis-related genes, whereas the low score group was enriched for immune response-associated genes. The TIIC signature score was significantly correlated with tumor-infiltrating immune cells, various metabolic characteristics, and chromosomal instability, and it effectively predicted immunotherapy response across diverse cancer types. CONCLUSION: The findings of this study highlighted the promise of the TIIC signature score in forecasting the outcomes for CRC patients. Additionally, it emphasized its utility in predicting the effects of immunotherapy, thereby enhancing our comprehension of the intricacies within the tumor microenvironment. Further research needs to concentrate on assessing the clinical utility of the TIIC signature score while also confirming its relevance across various populations and treatment contexts.