基于RNA-seq的结直肠癌分子与微环境分层：一个韩国真实世界队列研究

Colorectal cancer (CRC) continues to represent a major cause of cancer-related death, emphasizing the limitations of traditional staging systems in capturing clinically relevant biological heterogeneity. To enable more precise molecular stratification, we analyzed comprehensive transcriptomic profiles from 476 fresh-frozen colorectal specimens, including tumor and adjacent normal tissues, obtained from a real-world Korean cohort treated at Samsung Medical Center (SMC). By integrating RNA sequencing-derived expression data with clinicopathological variables and survival outcomes, we observed distinct transcriptional pathways separating tumor and normal tissues. Tumor samples displayed coordinated activation of pathways governing cell cycle progression, extracellular matrix (ECM) organization, and DNA damage response (DDR), whereas normal tissues were predominantly characterized by metabolic pathway activity. The enrichment of DNA repair-related pathways in tumors prompted focused evaluation of homologous recombination signaling, and homologous recombination deficiency (HRD) was quantified using a transcriptome-based scoring framework. Computational analyses incorporating tumor microenvironment (TME) signatures resolved four biologically distinct molecular subgroups. Tumors enriched for fibroblasts or combined immune-fibroblast features showed strong correspondence with the stromal-dominant CMS4 subtype, while immune-dominant tumors aligned preferentially with CMS1. Deconvolution of cellular composition and immune profiling revealed pronounced subgroup-specific differences in immune infiltration, stromal activation, and HRD burden. Importantly, these transcriptome-defined subgroups demonstrated significant divergence in clinical outcomes, with immune-fibroblast-enriched tumors exhibiting the most unfavorable prognosis independent of pathological stage. Moreover, combined assessment of HRD scores and CRC-relevant immune gene expression enabled further refinement of patient risk stratification into biologically coherent and clinically meaningful categories. In summary, this large-scale real-world transcriptomic analysis of Korean CRC patients delineates clinically actionable molecular subgroups defined by distinct TME architecture and homologous recombination status, underscoring the value of integrating RNA-seq-derived HRD and microenvironmental profiling to improve prognostic assessment and guide personalized therapeutic strategies in CRC.