AI Business LensTHE BUSINESS OF AI, FOR PEOPLE WHO TEACH IT OR LEARN FROM IT
Medical Xpress · September 10, 2026

AI-designed proteins outperform existing CAR T designs against BCMA tumors in mice

Biomedical EngineeringMedicineBiology
THE AI ANGLE
Designing and screening de novo CAR T protein binders from scratch

Researchers at Memorial Sloan Kettering used generative AI and a custom neural network, CARPNN, to design de novo protein binders for CAR T cells from scratch rather than relying on repurposed antibody fragments. In mouse models of multiple myeloma, their synthetic binder targeting BCMA significantly outperformed an FDA-approved clinical design by achieving superior tumor control. The team also evaluated failed designs against targets like CD19 to uncover fundamental biophysical rules, linking factors like excessive positive charge and high alanine content to premature activation and poor performance.

THE TEACHING ANGLE
Students can examine the biophysical failure modes revealed by the AI models—such as excessive positive charge causing premature T-cell activation or algorithm biases toward alanine—to debate the challenges of translating de novo computational designs into viable immunotherapies.

Read the original at medicalxpress.com   Generate teaching or study materials

Instructors get discussion guides, assignments, and mini-cases. Students and readers get a plain summary, class prep, and an exercise. All built from the full article. Three are free with an account.

More in Biomedical Engineering