Machine learning used to engineer molecular interactions

thank you. Listen to this article in the player above. ✖

In 2019, scientists at the Faculty of Engineering and Life Sciences’ Protein Design and Immunology Engineering Institute (LPDI), led by Bruno Correia, developed MaSIF. MaSIF scans millions of protein surfaces within minutes to analyze their structure and functional properties. The researchers’ ultimate goal was to computationally design protein interactions by finding the best match between molecules based on their surface chemical and geometric ‘fingerprints’.

Four years later they have achieved just that. In a paper published in Nature, They report creating entirely new proteins, called binders, designed to interact with four therapeutically relevant protein targets, including the SARS-CoV-2 spike protein.

Design a perfect molecular match

The ability to control protein-protein interactions is of great interest in biology and biotechnology because the physical interactions between proteins affect everything from cell signaling and growth to immune responses. increase. Textbook depictions of protein binding may seem as simple as putting together puzzle pieces, but the reality is much more complex.

“The pieces of the puzzle are two-dimensional, but on the protein surface we see multiple dimensions, such as chemical composition, such as the interaction of positive and negative charges. Shape complementarity, curvature, etc.,” says the LPDI Ph.D. student and co-author Anthony Marchand explains.

“The idea that all things in nature that combine are complementary (e.g., positive charges combine with negative charges) is a longstanding idea in the field and captured in computational frameworks.”

To design new protein binders, researchers used MaSIF to ‘fingerprint’ the protein surface and identify the complementary surfaces of key protein target sites from a database of fragments. We then digitally grafted the fragments onto a larger protein scaffold and selected the resulting binders predicted to interact best with the target. After synthesizing and testing these selected binders in the lab, the researchers were able to confirm their computationally generated hypotheses.

“The fact that we can design novel site-specific protein binders in just a few months makes this method very interesting for therapeutics. It’s not just a tool, it’s a pipeline,” says Marchand. .

“Right from your computer”

Researchers added the SARS-CoV-2 spike protein to the list as they were developing protein binders for three major cancer immunotherapy targets when the COVID pandemic hit. Using their approach, the four of his Binders they created showed great affinity for their targets.

The success rate of MaSIF, coupled with its speed and ability to create high-quality, site-specific designs, demonstrate its therapeutic potential. The ability to generate precise protein binders very quickly, for example, as in the case of the SARS-CoV-2 spike protein, could be of great advantage for epidemiological applications. Marchand also believes the pipeline could facilitate the development of chimeric antigen receptor (CAR-T) proteins. This can be manipulated to allow the patient’s immune cells to target cancer cells.

“Further advances in machine learning methods will help us improve our methods, but our task today is to rapidly design protein-based therapeutics directly from the computer to benefit patients. We are already providing strategies for developing innovative therapies that bring about

reference: Gainza P, Wehrle S, Van Hall-Beauvais A, et al. Novel designs of protein interactions with learned surface fingerprints. Nature2023;617(7959):176-184. Doi: 10.1038/s41586-023-05993-x

This article is reprinted from the following sources: Note: The length and content of the material may have been redacted. Please contact the citation source for details.