Sensitive detection of synthetic response to cancer immunotherapy driven by gene paralog pairs

Area of Science:

• Genomics and Bioinformatics
• Cancer Immunology
• Computational Biology

Background:

• Cancer immunotherapies like immune checkpoint blockade (ICB) and chimeric antigen receptor T-cell (CAR-T) therapy have transformed cancer treatment but face challenges with patient response and relapse.
• Current strategies using CRISPR screens to identify single gene targets for enhancing T-cell function have limited success due to cancer cells' complex, multi-gene immunosuppressive pathways.
• Paralogs, genes from common ancestors with similar functions, present an underexplored area for enhancing immunotherapy, despite their known roles in cancer cell survival and complex phenotypic effects.

Purpose of the Study:

• To develop a computational approach for identifying cancer-intrinsic paralog pairs that can synergistically enhance T-cell-mediated tumor destruction.
• To create an ensemble learning model for predicting paralog pairs likely to improve immunotherapy efficacy.
• To experimentally validate the functional significance of predicted paralog pairs in enhancing cancer immunotherapy.

Main Methods:

• Utilized a computational approach employing sgRNA sets enrichment analysis to identify potential cancer-intrinsic paralog pairs.
• Developed an ensemble learning model (XGBoost classifier) integrating gene characteristics, sequence/structural similarities, protein-protein interaction networks, and gene coevolution data for prediction.
• Experimentally validated predicted paralog pairs through double knockout (DKO) experiments, comparing their efficacy against single gene knockouts (SKOs).

Main Results:

• Identified novel paralog pairs with the potential to synergistically enhance T-cell-mediated tumor destruction.
• The ensemble learning model successfully predicted paralog pairs likely to improve immunotherapy efficacy based on integrated features.
• Experimental validation confirmed the functional significance of identified paralog pairs, demonstrating enhanced immunotherapy effects compared to single gene targeting.

Conclusions:

• The developed computational and ensemble learning approach provides a sensitive method for identifying previously undetected paralog pairs that can enhance cancer immunotherapy.
• Targeting specific paralog pairs offers a promising strategy to overcome limitations of single-gene targeting and improve patient response to existing immunotherapies.
• This research opens new avenues for discovering synergistic targets to boost the efficacy of cancer immunotherapies, even when individual gene modulation has limited impact.