A I-designed proteins may help spot cancer
AI model designs molecular sensors that detect cancer-linked enzymes, enabling a simple at-home urine test.
A collaboration between MIT and Microsoft Research has yielded a breakthrough in early cancer detection, using AI to design molecular sensors detectable in urine. Led by MIT's Sangeeta Bhatia and Microsoft's Ava Amini, the team developed an AI model that designs short, targeted proteins (peptides) which react with specific enzymes called proteases that are overactive in cancer cells. When these AI-designed peptides, coated on nanoparticles, encounter cancer-linked proteases in the bloodstream, they are snipped off and form reporter molecules that are eventually excreted. This provides a clear, non-invasive signal of disease, a significant leap from previous ambiguous, trial-and-error peptide design methods.
The technical advance lies in the AI's ability to optimize peptides for high sensitivity and specificity to particular cancer proteases, creating a powerful diagnostic signal. The implications are profound: Bhatia's lab is now working with the U.S. Advanced Research Projects Agency for Health (ARPA-H) to develop an at-home test kit with the potential to screen for 30 different early-stage cancers from a urine sample. Beyond diagnostics, these precisely engineered peptides could also be incorporated into future targeted cancer therapeutics. This research, over a decade in the making, demonstrates how AI is accelerating biomedical discovery by moving from brute-force experimentation to rational, criteria-driven design of biological tools.
- AI model designs short protein sensors (peptides) targeted by cancer-linked proteases.
- Protease activity cleaves peptides, creating reporter molecules detectable in a simple urine test.
- Team collaborating with ARPA-H on an at-home kit to potentially detect 30 early cancer types.
Why It Matters
Enables non-invasive, early cancer screening at home, moving detection from clinics to daily life with AI-driven precision.