AI-guided CART platform targets two relapse drivers: poor persistence and antigen escape

Feb. 13, 2026 —

A research team led by Tanveer Ahmad reports a next-generation CAR T engineering strategy aimed at two leading causes of relapse in hematologic malignancies: poor CAR T-cell persistence and tumor antigen escape. The work was published by a multi-institution collaboration anchored at the Multidisciplinary Centre for Advanced Research and Studies, Jamia Millia Islamia, with contributing partners spanning academia, hospitals, and industry.

The study integrates AI-guided CAR design with targeted protein degradation to improve durability. Researchers first built a large in-silico library of CAR constructs and paired it with in vitro screening to train CARMSeD, a predictive model that flags CAR designs prone to self-activation and dysfunction. Using this workflow, the team optimized bispecific CD20/CD19 CAR T cells, reporting improved persistence and enhanced anti-tumor activity.

To further extend functional longevity, the platform adds a PROTAC-based module that selectively degrades AKT3, shifting cells toward a fitter, memory-like state. The authors report this intracellular tuning supports FOXO4-linked mitochondrial fitness, increases central memory differentiation, and reduces mTOR signaling—changes associated with reduced exhaustion and improved persistence.

Finally, to counter antigen-negative relapse, the researchers extended the approach into a trispecific CAR T concept that co-expresses a secretable CD3/CD22 bispecific engager, enabling tumor killing even in CD19/CD20-negative settings. The strategy demonstrated activity across patient-derived leukemia samples and was also evaluated in solid tumor models, highlighting potential broader applicability.