Colorectal melanoma (CRM) is a rare malignant tumor with severe complications, and there is currently a lack of systematic research. We conducted a study that combined proteomics and mutation data of CRM from a cohort of three centers over a 16-years period (2005-2021). The patients were divided into a training set consisting of two centers and a testing set comprising the other center. Unsupervised clustering was conducted on the training set to form two molecular subtypes for clinical characterization and functional analysis. The testing set was used to validate the survival differences between the two subtypes. The comprehensive analysis identified two subtypes of CRM: immune exhausted C1 cluster and DNA repair C2 cluster. The former subtype exhibited characteristics of metabolic disturbance, immune suppression, and poor prognosis, along with APC mutations. A machine learning algorithm named Support Vector Machine (SVM) was applied to predict the classification of CRM patients based on protein expression in the external testing cohort. Two subtypes of primary CRM with clinical and proteomic characteristics provides a reference for subsequent diagnosis and treatments.
Keywords: Machine learning; Molecular subtype; Multicenter; Primary colorectal melanoma; Proteomics.
Copyright © 2024 The Author(s). Published by Elsevier B.V. All rights reserved.