Biology-aware mutation-based deep learning for outcome prediction of cancer immunotherapy with immune checkpoint...

Area of Science:

• Oncology
• Computational Biology
• Genomics

Background:

• Predicting patient response to immune checkpoint inhibitors (ICI) is challenging due to variable outcomes.
• Gene mutations are crucial for treatment success, but existing models lack biological context.

Purpose of the Study:

• To develop and validate a novel mutation-based deep learning model for survival analysis in patients treated with ICI.
• To incorporate biological knowledge into a predictive framework for enhanced accuracy and interpretability.

Main Methods:

• A deep learning model was developed using gene mutation data from 1571 cancer patients treated with ICI.
• The model incorporated gene pathways and protein interactions for a biology-aware approach.
• Performance was evaluated using survival analysis and compared against the Cox-PH model.

Main Results:

• The proposed deep learning model achieved an average concordance index of 0.59 ± 0.13 across nine cancer types.
• This performance surpasses the gold standard Cox-PH model (0.52 ± 0.10).
• The model demonstrated interpretability, aiding in patient stratification and biomarker discovery.

Conclusions:

• Mutation-based deep learning, incorporating biological knowledge, offers superior survival prediction for ICI therapy.
• The biology-aware model enhances patient stratification and provides insights into novel gene biomarkers.
• This approach advances the understanding and application of immunotherapy.