A I-designed proteins may help spot cancer

Researchers at MIT and Microsoft have used artificial intelligence to create molecular sensors that could detect early signs of cancer via a urine test.

The researchers developed an AI model to design short proteins that are targeted by enzymes called proteases, which are overactive in cancer cells. Nanoparticles coated with these proteins, called peptides, can give off a signal if they encounter cancer-­linked proteases once introduced into circulation: The proteases will snip off the peptides, which then form reporter molecules that are excreted in urine.

Sangeeta Bhatia, SM ’93, PhD ’97, a senior author of a paper on the work with her former student Ava Amini ’16, a principal researcher at Microsoft Research, led the MIT team that came up with the idea of such particles over a decade ago. But earlier efforts used trial and error to identify peptides that would be cleaved by specific proteases, and the results could be ambiguous. With AI, peptides can be designed to meet specific criteria.

“If we know that a particular protease is really key to a certain cancer, and we can optimize the sensor to be highly sensitive and specific to that protease, then that gives us a great diagnostic signal,” Amini says. 

Bhatia’s lab is now working with ARPA-H on an at-home kit that could potentially detect 30 types of early cancer. Peptides designed using the model could also be incorporated into cancer therapeutics.