MSIpred: a python package for tumor microsatellite instability classification from tumor mutation annotation data...

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Microsatellite instability (MSI) is a hallmark of deficient mismatch repair (MMR) and a prognostic indicator in various cancers.
• Conventional MSI detection via MSI-PCR is labor-intensive, costly, and time-consuming.

Purpose of the Study:

• To develop an automated, machine learning-based tool, MSIpred, for accurate MSI classification.
• To provide a robust and efficient alternative to conventional MSI detection methods.

Main Methods:

• Developed MSIpred, a Python package utilizing Support Vector Machine (SVM) machine learning.
• Extracted 22 features from mutation annotation format (MAF) data of paired tumor-normal exome sequencing.
• Trained the SVM classifier on MAF data from 1074 tumors across four types.

Main Results:

• Achieved ≥98% accuracy and an ROC AUC of 0.967 on an independent test set of 358 tumors.
• Identified potential misclassifications in conventional MSI-PCR, highlighting MSIpred's improved accuracy.
• Validated MSIpred's strong performance on non-TCGA datasets, demonstrating its pan-tumor applicability.

Conclusions:

• MSIpred is a robust, automated tool for pan-tumor MSI classification.
• MSIpred serves as a valuable complementary diagnostic to MSI-PCR, enhancing MSI detection efficiency and accuracy.