Pathway-based molecular subtyping identifies LAMP5 as an EMT-associated prognostic target in colorectal cancer
Colorectal cancer (CRC) is a highly heterogeneous malignancy with substantial variability in clinical outcomes. Although established molecular classification systems have improved CRC stratification, pathway-level subtyping remains underexplored. A framework integrating pathway-based molecular subtype discovery, multi-cohort validation, single-cell projection, and mechanistic target identification is still lacking. Transcriptomic data from TCGA CRC samples were transformed into pathway activity profiles using gene set variation analysis (GSVA), followed by consensus clustering to define pathway-based molecular subtypes. Nearest template prediction (NTP) was applied across 17 independent cohorts to validate subtype robustness. Three CRC single-cell RNA-seq datasets were integrated and analyzed using the single-cell phenotype-associated subpopulation identifier (scPAS) to map subtype-associated signals at cellular resolution. Multi-cohort Cox regression identified prognostic genes, and LAMP5 was further assessed through clinicopathological correlation, pathway enrichment analyses, and experimental validation in vitro and in vivo . Four pathway-based molecular subtypes were identified, among which C4 had the worst prognosis. The subtype classification was reproducible across 17 external cohorts. Single-cell analysis suggested that the aggressive subtype-associated phenotype could be projected onto distinct cell populations. LAMP5 emerged as the most consistent unfavorable prognostic gene across cohorts and was associated with advanced clinicopathological features. Functional analyses linked LAMP5 to TGFβ signaling, epithelial-mesenchymal transition (EMT), angiogenesis, invasion, and metastasis. Experimental results supported the pro-tumorigenic role of LAMP5 in CRC. This study establishes a pathway-based molecular classification framework for CRC and identifies LAMP5 as a robust prognostic biomarker and candidate therapeutic target associated with the EMT pathway.