Multi-omics factor analysis identifies the Tensin 1-Fermitin family homologue 2–Fibronectin 1–Integrin signaling axis as a prognostic determinant in colorectal cancer

Colorectal cancer (CRC) remains a leading cause of cancer-related mortality worldwide, with its heterogeneous molecular profiles complicating prognosis prediction. Multi-omics profiling of clinical CRC specimens is particularly valuable for uncovering prognostic variables; however, such integrated studies remain scarce. Here, we applied the Multi-Omics Factor Analysis v2 framework to multi-omics data (mutations, miRNA, RNA, proteomics, phospho-proteomics) from the CPTAC-2 cohort and constructed a comprehensive CRC model. This approach identified a survival-associated latent variable, which we term the CRC Prognostic Latent Factor (CPLF). The prognostic relevance of CPLF was then rigorously validated across three independent multi-omics cohorts, encompassing 579 patients with CRC. CPLF primarily reflects extracellular matrix deposition in the tumour microenvironment, with Tensin 1 (TNS1) and Fermitin family homologue 2 (FERMT2) as the highest-weighted features. Single-cell and spatial transcriptomic analyses, supplemented by immunohistochemistry, localized CPLF-associated gene expression predominantly to myofibroblasts. Functionally, knockdown of Tensin 1 or FERMT2 in fibroblasts attenuated tumour growth in vivo. Consistently, these top-weighted components of CPLF upregulated fibronectin 1 (FN1) expression in myofibroblasts, thereby activating integrin signaling in cancer cells and enhancing tumour progression. Altogether, our findings unveil CPLF as a data-inherent multi-omics prognostic factor and establish the CPLF/FN1/integrin axis as a key pathway mediating myofibroblast-cancer cell crosstalk in CRC progression.