MSIpred

MSIpred classifies tumor microsatellite instability (MSI) status using a support vector machine applied to somatic mutation features extracted from mutation annotation format (MAF) files generated from paired tumor-normal exome sequencing.

Key Features:

• Input data: Accepts mutation annotation format (MAF) files derived from paired tumor-normal exome sequencing.
• Feature extraction: Computes 22 distinct features that characterize the tumor somatic mutational load.
• Classification algorithm: Uses a support vector machine (SVM) classifier for MSI status prediction.
• Training data: Classifier trained on MAF data from 1,074 tumors across four tumor types.
• Performance: Reported overall classification accuracy ≥98% and area under the ROC curve of 0.967 on independent testing sets.
• Cross-dataset robustness: Demonstrated performance on non-TCGA datasets indicating applicability across multiple tumor cohorts.
• Automated MSI calling: Produces MSI status predictions directly from somatic mutation features extracted from MAF files.

Scientific Applications:

• Diagnostic complement: Serves as a computational complement to MSI-PCR for determining MSI status from sequencing-derived variant data.
• Prognostic assessment: Enables MSI status-based stratification relevant to tumor prognosis studies.
• Treatment guidance: Provides MSI status information relevant to research and clinical decision-making for personalized treatment planning.

Methodology:

Computes 22 somatic mutational-load features from MAF files derived from paired tumor-normal exome sequencing and applies an SVM classifier trained on MAF data from 1,074 tumors across four tumor types.