Artificial Intelligence Framework Creates Tailored Proteins Targeting Specific Drug Molecules

A Dana-Farber Cancer Institute team led by Nicholas Polizzi, PhD, has developed an artificial intelligence method that designs proteins to recognize and hold specific drug molecules. These custom proteins could eventually help scientists make cancer medicines safer, more precise and easier to monitor.

The method, called neural iterative selection-expansion (NISE), works as a computer-guided design cycle. It proposes a protein, predicts how well the protein and drug fit together and then improves the design. This allows researchers to start with a drug of interest and quickly generate a small number of promising proteins for laboratory testing, rather than screening thousands of possibilities by trial and error. In this study, nearly all of the proteins selected for testing bound the intended drug.

The researchers first designed a protein for exatecan, a potent cancer drug used in antibody-drug conjugates. The protein wrapped around the drug and protected a fragile part that normally breaks down. In the future, proteins like this might be developed to keep cancer drugs active, deliver them more precisely, measure where they are in the body or capture a drug that is released in the wrong place and causes side effects. As a second test, the team designed a protein that bound the blood thinner apixaban, showing that the method can be adapted to very different medicines. Patients taking apixaban may need its effects reversed quickly after serious bleeding or before emergency surgery. The designed protein could be a starting point for a future antidote that captures apixaban and neutralizes its blood-thinning effect. These potential uses have not yet been tested in animals or patients.

Significance

Many cancer medicines are powerful but difficult to control after they enter the body. A protein made to recognize one particular drug could potentially serve as a carrier, protective shell, sensor or sponge. Until now, creating such proteins has often required large experimental screens. NISE moves much of that search onto the computer and could give cancer researchers a faster way to build new tools around existing and future therapies. Although the technology is still at an early stage, it could ultimately support safer drug delivery, better measurement of drug exposure and new ways to limit toxicity.

Funding

The National Science Foundation, the National Institutes of Health and the Innovation Research Fund at Dana-Farber Cancer Institute