Computational identification of cross-reactive TCR epitopes with ARDitox

Area of Science:

• Immunology
• Computational Biology
• Artificial Intelligence

Background:

• Cellular immunotherapies using T cell receptors (TCRs) show therapeutic promise.
• Engineered TCRs can cause fatal off-target toxicity due to cross-reactivity with healthy tissues.
• Predicting TCR cross-reactivity is challenging, and mouse models miss human-specific epitopes.

Purpose of the Study:

• To develop and validate ARDitox, an in silico method for predicting TCR off-target toxicities.
• To assess ARDitox's ability to identify potentially harmful cross-reactive epitopes.
• To improve the safety of TCR-mediated immunotherapies.

Main Methods:

• Developed ARDitox, an AI-based computational immunology tool.
• Tested ARDitox on four TCRs targeting tumor-associated antigens, including known toxic and novel targets.
• Analyzed TCR cross-reactivity against known and predicted epitopes.

Main Results:

• ARDitox accurately confirmed previously identified immunotoxic epitopes.
• Identified a novel cross-reactive epitope for a glioma-targeting TCR (NLGN4X).
• The identified epitope would not have been detected using traditional mouse models.

Conclusions:

• ARDitox enables early and reliable identification of off-target epitopes.
• The computational platform supports preclinical evaluation of TCRs.
• ARDitox is crucial for developing safer TCR-mediated immunotherapies.