A Machine Learning-Based Scoring System to Identify High Immunoactivity Microsatellite Stability Tumors by...

Area of Science:

• Oncology
• Immunotherapy
• Computational Biology

Background:

• Microsatellite stable (MSS) colorectal cancers (CRCs) exhibit limited response to immune checkpoint inhibitors (ICIs) compared to microsatellite instability-high (MSI-H) CRCs.
• Some MSS CRCs show sensitivity to ICIs, but criteria for treatment selection are lacking.

Purpose of the Study:

• To analyze tumor-infiltrating lymphocyte (TIL) features in MSS CRCs.
• To develop a computational tool predicting MSS CRC similarity to MSI-H status based on multiple factors.

Main Methods:

• Collected and analyzed data from 188 CRC patients (MSI status, immune cells, clinical features, gene mutations).
• Developed an ensemble machine learning-based MSI-H score using extreme gradient boosting classifiers.
• Quantified similarity to MSI-H status using immune cell distributions, clinical features, and gene mutations.

Main Results:

• The MSI-H score accurately identified MSI-H-like MSS samples with TIL distributions similar to genuine MSI-H CRCs (mean Cohen κ of 0.40).
• No significant difference in TIL features was found between MSI-H-like MSS CRCs and MSI-H CRCs.
• Potential disparities between MSI-H-like MSS and MSS CRCs involve T regulatory cells and macrophage populations.

Conclusions:

• Some MSS CRCs exhibit high immunoactivity comparable to MSI-H CRCs.
• The MSI-H score quantifies MSS CRC similarity to MSI-H CRCs, aiding personalized immunotherapy.
• This metric offers a clinical reference for potential ICI targets in MSS CRCs.