New Method Tackles Inverse Protein Folding Problem

Researchers from the University of Sheffield and AstraZeneca have developed an artificial intelligence approach that aims to simplify the design of protein therapeutics. This new method, described in Nature Machine Intelligence, addresses the challenge of inverse protein folding, a complex process that is essential for creating proteins with specific functions.

Inverse protein folding involves determining which amino acid sequences will fold into a desired three-dimensional structure, enabling the protein to perform its intended function. This step is crucial in protein engineering, which plays a significant role in drug development by designing proteins that can bind to specific targets in the body. However, predicting how small changes in protein sequences will affect their structure is difficult, making the process challenging.

To improve accuracy, scientists have increasingly turned to machine learning models trained on large datasets of known protein sequences and structures. The new framework, called MapDiff, was developed by teams at the University of Sheffield, AstraZeneca, and the University of Southampton. In simulated tests, MapDiff outperformed current state-of-the-art AI methods for inverse protein folding, making more successful predictions about which sequences will fold into stable, functional structures.

According to Haiping Lu, corresponding author of the study, “This work represents a significant step forward in using AI to design proteins with desired structures. By learning how to generate amino acid sequences that are likely to fold into specific 3D structures, our method opens new possibilities for designing new therapeutic proteins, which can be used in various therapeutic applications.”

The study builds on previous collaborations between Sheffield computer scientists and AstraZeneca, including the development of DrugBAN, an AI tool for predicting drug-target interactions.