Abstract
Gastrointestinal (GI) cancers are a heterogeneous group of primary solid tumors, arising in GI tract from the esophagus to rectum. Matrix stiffness (MS) is a critical physical factor for cancer progression; however, its importance in tumor progression remains to be comprehensively recognized. Herein, we conducted a comprehensive pan-cancer analysis of MS subtypes across seven GI-cancer types. Using unsupervised clustering based on literature-derived MS-specific pathway signatures, the GI-tumor samples were divided into three MS subtypes, termed as the Soft, Mixed and Stiff. Then, distinct prognoses, biological features, tumor microenvironments and mutation landscapes among three MS subtypes were revealed. The Stiff tumor subtype was associated with the poorest prognosis, the most malignant biological behaviors, and the immunosuppressive tumor stromal microenvironment. Furthermore, multiple machine learning algorithms were used to develop an 11-gene MS-signature to identify the MS subtypes of GI-caner and predict chemotherapy sensitivity, which were further validated in two external GI-cancer cohorts. This novel MS-based classification on GI-cancers could enhance our understanding of the important role of MS in tumor progression, and may have implications for the optimization of individualized cancer management.