AI-guided CAR designs and targeted pathway modulation to enhance multi-antigen CAR T cell durability and overcome antigen escape
Cet article présente une stratégie de développement guidée par l'intelligence artificielle pour améliorer la persistance des lymphocytes CAR-T, élargir la couverture antigénique de ces derniers et garantir une efficacité thérapeutique durable
The persistence of CAR T cells and antigen escape remain major barriers to durable therapeutic success in hematologic malignancies. Our study integrates AI-guided design with targeted protein degradation to overcome these challenges. Utilizing an in-silico library of CAR constructs followed by an in vitro screening, we developed a predictive model, CARMSeD, which forecasts constructs prone to self-activation and dysfunction. Optimized bispecific CD20/CD19 CAR T cells demonstrate superior persistence and anti-tumor efficacy. To further improve durability, the platform incorporates a PROTAC-based module that selectively degrades AKT3, promoting FOXO4-driven mitochondrial fitness, central memory differentiation, and reduced mTOR signaling. We extended this strategy to develop a trispecific CAR T platform co-expressing a secretable CD3/CD22 bispecific engager, achieving potent tumor eradication even in CD19/CD20-negative malignancies demonstrates efficacy across patient-derived leukemia samples and solid tumor models. Together, our study introduces a next-generation AI-guided CAR T strategy that integrates structure-based optimization and intracellular modulation to improve persistence, broaden antigen coverage, and ensure durable therapeutic efficacy.
Nature Communications , article en libre accès, 2026