Researchers at Ecole Polytechnic Federale de Lausanne (EPFL) in Switzerland have developed a machine learning approach to scan millions of protein fragments and assess their structure and binding properties. Based on the protein’s surface chemistry and shape, the developed software can determine the ‘fingerprint’ of each protein and predict how the protein will bind to different protein fragments. Researchers are now using that approach to design new protein ‘binders’ specifically crafted to bind proteins of therapeutic interest, such as the SARS-CoV-2 spike protein. The technique could allow researchers to create a series of therapeutic proteins very quickly, which could be especially useful in time-sensitive situations such as future pandemics.
Many factors influence how and whether proteins bind to each other, making it difficult to predict using only the human brain. However, identifying protein fragments that can interact with therapeutic protein targets in the body could have enormous clinical benefit in the treatment of various diseases. Thankfully, computers are well adapted to tasks that involve a daunting amount of detailed data and its complex sorting. This state-of-the-art approach involves deep learning, allowing us to assess the subtle interactions between the various factors that influence coupling.
“While the pieces of the puzzle are two-dimensional, we are looking at protein surfaces in multiple dimensions, such as chemical composition, such as interactions between positive and negative charges, shape complementarity, curvature,” he says. Researcher Anthony Marchand, who was involved in the study, said: “The idea that everything that binds in nature is complementary (e.g., a positive charge binds with a negative charge) is a longstanding idea in the field, and we captured it in our computational framework. rice field.”
So far, researchers have used the system to create a series of protein “binders” that can bind to therapeutic targets such as the SARS-CoV-2 spike protein. This involved using a deep learning approach to create a protein ‘fingerprint’ and then scanning a database of protein fragments to find fragments predicted to bind well with that fingerprint. . We then tested the likelihood that the fragments with the best predicted avidity would actually bind to the target, first through digital simulation, and finally in the laboratory after the fragment was synthesized.
“The fact that we can design novel site-specific protein binders in just a few months makes this method very interesting for therapy. It’s not just a tool, it’s a pipeline,” said Marchand. Told. “Further advances in machine learning methods will help us improve our method, but our efforts today are focused on rapidly designing protein-based therapeutics directly from the computer to benefit patients. We are already providing strategies to develop innovative therapies.”
study abroad Nature: de novo Designing protein interactions using learned surface fingerprints
Via: EPFL
