GDAN Tumor Molecular Pathology

The Genomic Data Analysis Network (GDAN) Tumor Molecular Pathology (TMP) project developed machine-learning models that assign TCGA-defined molecular subtype labels to cancer samples from other studies and clinical trials. The study trained five modeling approaches on 8,791 TCGA tumors spanning 106 subtypes across 26 cancer cohorts, and validated selected classifiers using external datasets.

The TMP model repository provides containerized models for gene expression, DNA methylation, miRNA, somatic mutation, and copy-number data, along with tools and instructions for preparing data and running predictions. Read the Cancer Cell paper for the study and its results.