TMScal: A Clinical Decision Support Tool for Prognostic Stratification of Microsatellite-Stable Cancers

PURPOSE Microsatellite-stable (MSS) cancers comprise 70–85% of patients but lack prognostic biomarkers to guide clinical decision-making. Current metrics—tumor mutational burden (TMB), copy number variation (CNV), and microsatellite instability (MSI)—measure the quantity of genomic instability but discard sequence-context information that encodes repair pathway activity. We developed TMScal, a clinical decision support tool implementing the Trinucleotide Mutation Spectrum (TMS) framework, and validated its prognostic utility across three independent cohorts totaling 4,913 MSS patients.METHODS TMScal is an R package that computes TMS scores from sequencing data by analyzing the proportional distribution of somatic mutations across 192 complete trinucleotide-by-base-change pathways. The tool features automatic format detection, Elastic Net Cox regression with automated hyperparameter optimization, and single-patient prediction with comprehensive clinical reporting. We validated TMScal using a 70:30 split strategy across three cohorts: SYSUCC-COADREAD (discovery, WES, N=662), TCGA-UCEC (validation, WES, N=400), and MSK-IMPACT COADREAD (clinical validation, targeted panel, N=3,851). One hundred repeated splits were performed to assess robustness.RESULTS TMScal stratified MSS patients into three distinct risk groups in independent validation sets: SYSUCC-COADREAD (HR=2.54, 95% CI: 1.16–5.57, P=0.0203, C-index=0.604); TCGA-UCEC (HR=2.93, 95% CI: 0.90–9.54, P=0.0739, C-index=0.637, showing a trend toward significance); and MSK-IMPACT COADREAD (HR=1.34, 95% CI: 1.06–1.70, P=0.0155, C-index=0.555). One hundred repeated splits confirmed result stability, with mean HRs of 2.09±0.81 (SYSUCC), 2.42±1.06 (UCEC), and 1.28±0.13 (MSK). TMScal generates clinical reports in under five minutes, requiring no additional tissue or sequencing beyond routine clinical testing.CONCLUSION TMScal is an immediately deployable, clinical decision support tool that provides robust prognostic stratification for MSS cancer patients using routine clinical sequencing data. The tool is freely available at https://github.com/cxue/TMScal under Apache License 2.0.HOW TO ACCESS/USE TMScal is implemented in R (≥ 4.2.0). The package includes comprehensive documentation, vignettes, pre-trained models for multiple cancer types (colorectal and endometrial), and example datasets for testing. Installation and usage instructions are provided in the package README and vignettes.