AI Predicts Protein-Fragment Inhibitors for Targeted Therapies

A new method, FragFold, uses artificial intelligence to predict protein fragments that can bind to and inhibit full-length proteins. Developed in the Department of Biology, the tool leverages AlphaFold, an AI model known for predicting protein folding and interactions. Researchers confirmed that over half of FragFold's predictions for binding or inhibition were accurate, even without prior structural data. This approach could be applied to proteins with unknown functions or structures. The researchers explored fragments of FtsZ, a protein key for cell division, identifying new binding interactions. Deep mutational scanning revealed key amino acids responsible for inhibition, with some mutated fragments proving more potent than full-length sequences. FragFold opens possibilities for manipulating protein function and creating new tools for studying cell biology and treating diseases.