Sequence and structural determinants of efficacious de novo chimeric antigen receptors

Advances in generative protein design using artificial intelligence (AI) have enabled the rapid development of binders against heterogeneous targets, including tumor-associated antigens. Despite extensive biochemical characterization, these novel protein binders have had limited evaluation as agents in candidate therapeutics, including chimeric antigen receptor (CAR) T cells. Here, we synthesize variable generative protein design workflows to screen 1,487 novel protein binders targeting BCMA, CD22, and CD19 for efficacy in scalable protein binding and T cell assays. We identify three main challenges that hinder the utility of de novo protein binders as CARs, including tonic signaling, occluded epitope engagement, and off-target activity. We develop computational and experimental heuristics to overcome these limitations, including screens of sequence variants for individual parental structures, that restore on-target CAR activation while mitigating liabilities. Together, our framework accelerates the development of AI-designed proteins for future preclinical therapeutic screening, helping enable a new generation of cellular therapies.