Optimising CAR T Cell Design

Chimeric antigen receptor (CAR) T-cell therapies show promise for treating cancers and other diseases, but designing effective CAR constructs remains a major challenge. It requires testing possible combinations of intracellular costimulatory motifs; short stretches of genetic or protein sequences. These motif configurations can strongly influence how the T cell behaves. Laboratory validation for each design is expensive, labour-intensive, and slow, meaning only a small fraction of possibilities can realistically be explored.

Quantum machine learning methods, such as Projected Quantum Kernels (PQK) and Quantum Convolutional Neural Networks (QCNN), can learn patterns linking motif configurations to cytotoxicity (the ability of CAR T cells to kill cancer cells) from a limited set of measured CAR T constructs. Both approaches can highlight untested designs most likely to perform well, focusing lab work on only the most promising candidates.

This could lead to more accurate and efficient predictions of therapeutic performance, which could reduce the number of required laboratory experiments and hence accelerate the development of more effective CAR T-cell therapies. It could also support the exploration of novel design combinations that may be overlooked using classical approaches. Over time, this could contribute to more targeted, robust, and cost-effective immunotherapies for a wider range of diseases.