Network-based machine learning approach to predict immunotherapy response in cancer patients

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Immune checkpoint inhibitors (ICIs) have transformed cancer therapy, yet patient response rates remain limited (~30%) in solid tumors.
• Existing biomarkers for ICI response often lack predictive power, hindering personalized treatment strategies.

Purpose of the Study:

• To develop and validate a novel machine learning (ML) framework, termed NetBio, for identifying robust biomarkers of ICI treatment response.
• To enhance the accuracy of predicting patient outcomes in response to immunotherapy across diverse cancer types.

Main Methods:

• Curated transcriptomic and clinical outcome data from over 700 ICI-treated patient samples.
• Employed network-based analyses within an ML framework to identify predictive biomarkers.
• Validated NetBio predictions against conventional ICI-response biomarkers.

Main Results:

• NetBio demonstrated accurate prediction of ICI treatment response in melanoma, gastric cancer, and bladder cancer.
• The NetBio framework outperformed traditional biomarkers, including ICI targets and tumor microenvironment markers.
• Network-based biomarker identification offers superior predictive capability for immunotherapy response.

Conclusions:

• The NetBio framework provides an effective method for selecting immunotherapy response biomarkers.
• This network-based ML approach facilitates robust predictions, paving the way for precision oncology.
• NetBio represents a significant advancement in predicting patient response to immune checkpoint inhibitors.