Integrating transcriptomics and machine learning for immunotherapy assessment in colorectal cancer

Area of Science:

• Oncology
• Immunology
• Bioinformatics

Background:

• Colorectal cancer (CRC) is a leading cause of cancer death globally.
• Immunotherapy has improved CRC outcomes, but patient intolerance is a significant challenge.
• The tumor microenvironment (TME) critically influences CRC progression and treatment response.

Purpose of the Study:

• To classify colorectal cancer (CRC) into distinct tumor microenvironment subtypes (TMESs) using transcriptomic data.
• To identify a transcriptomic signature predictive of immunotherapy response in CRC.
• To investigate the mechanistic interplay between TMESs, immunotherapy, and clinical outcomes.

Main Methods:

• Utilized non-negative matrix factorization to categorize 2595 CRC samples from the Gene Expression Omnibus database into TMESs.
• Employed machine learning techniques to identify an immunotherapy-specific signature.
• Analyzed the relationship between TMESs, activated dendritic cells, CD8+ T cells, and immunotherapy response.

Main Results:

• Identified five distinct TMESs (TMES1-TMES5) in CRC, each with unique immunotherapy response profiles.
• TMES2 showed the poorest prognosis, while TMES3 indicated a superior outcome.
• Activated dendritic cells, in conjunction with CD8+ T cell activation, were found to enhance immunotherapy response rates.

Conclusions:

• The five identified TMESs offer a framework for prognostic evaluation in CRC.
• These TMESs can guide personalized immunotherapy strategies for improved clinical decision-making.
• Understanding TME subtypes is crucial for optimizing CRC treatment and patient outcomes.