MmCMS

MmCMS performs consensus molecular subtyping of mouse colorectal cancer (CRC) models to map their transcriptional profiles to human CRC consensus molecular subtypes (CMS1–CMS4).

Key Features:

• Dual-Species Classification: Classifies mouse CRC models into human-equivalent CMS1–CMS4 subtypes to enable cross-species comparison of tumor molecular phenotypes.
• Comprehensive Methodology: Uses transcriptional data from established human and mouse CRC tumor collections and applies machine learning algorithms including random forest and nearest template prediction to develop classifiers.
• Multiple Classification Approaches: Implements three classifiers—MmCMS-A (gene-level classifier), MmCMS-B (ontology-based classifier using gene ontology collections), and MmCMS-C (pathway-level classifier integrating multiple biological and histological signaling cascades).
• Performance Evaluation: Evaluates classifier performance using the TCGA human CRC cohort (n=577) and mouse tumors (n=57 across 8 genotypes), reporting stronger detection of stromal-rich CMS4-like biology and limitations of MmCMS-A for epithelial-like CMS2/3 subtypes while identifying MmCMS-C as the optimal pathway-level approach.

Scientific Applications:

• Model selection for preclinical testing: Enables selection of mouse models that recapitulate specific human CRC subtypes for experimental and drug testing.
• Translational relevance assessment: Supports assessment of how well genetically engineered or experimental mouse tumors mirror human CRC molecular characteristics.
• Cross-species target and pathway exploration: Facilitates investigation of conserved therapeutic targets, signaling pathways, and treatment strategies across human and mouse CRC biology.

Methodology:

Uses transcriptional data from human and mouse CRC tumor collections; trains classifiers using machine learning methods including random forest and nearest template prediction; evaluates performance on the TCGA cohort (n=577) and on mouse tumors (n=57 across 8 genotypes), with the MmCMS-C classifier integrating multiple biological and histological signaling cascades.