This study introduces a machine learning model that uses RNA sequencing data to predict microsatellite instability (MSI) in cancer patients. The model employs a logistic regression algorithm to classify tumors as MSI or microsatellite stable (MSS) based on the mutational profiles of microsatellite hotspots. Analysis across six datasets comprising 486 samples demonstrated high accuracy, with an ROC AUC ranging from 0.85 to 1.00. This RNA-Seq-based approach provides a robust, tissue-agnostic tool for MSI detection, potentially improving the identification of patients likely to respond to immune checkpoint inhibitors.

Published at SITC 2023.