Computer-designed immune receptor makes engineered T cells better cancer killers

Researchers at Bar-Ilan University and the Weizmann Institute of Science have used computer-based protein design to build a stronger version of the receptor that T cells use to recognise disease. The study, published in Science Advances and announced on 24 September 2026, was led by Prof. Cyrille Cohen, head of Bar-Ilan's immunotherapy laboratory, together with Prof. Sarel Fleishman of the Weizmann Institute, whose group develops protein-design software.

A growing class of cancer treatments works by giving a patient's own T cells a new receptor, known as a TCR, that recognises a target on tumour cells. One weakness is that the added receptor chains can mispair with the cell's own natural chains, producing hybrid receptors that work poorly or not at all. Using the Fleishman lab's PROSS and FuncLib design tools, the team rewrote the constant region of the human TCR to create a scaffold they call SET, for Structurally Enhanced TCR, which pairs more reliably with its intended partner.

T cells carrying SET produced substantially stronger immune responses and killed cancer cells more effectively in laboratory tests than cells with the original receptor. In mice carrying human tumours, tumours treated with SET cells were about 35 percent smaller than untreated controls after 83 days, and all of the treated mice were still alive at day 127, compared with fewer than half of the controls. The same set of design changes also improved receptors aimed at melanoma-related targets and at proteins from the Epstein-Barr virus and SARS-CoV-2, suggesting it could be applied broadly rather than tailored to each therapy.

The first author was Dr. Maria Radman of Bar-Ilan, with Dr. Esther Shmuel-Revy carrying out much of the laboratory work and visiting Australian researcher Dr. Jake Parker contributing to the computational design. The results are preclinical, and the safety and effectiveness of SET-based cells in patients have yet to be tested.