How Artificial Intelligence Promotes Structural Proteomics

This article is based on research findings that have not yet been peer-reviewed. Results are therefore considered preliminary and should be interpreted as such. Read more about the role of the peer review process in research here† For more information, please contact the said source.

Understanding the formation of protein complexes is crucial in drug design and development of therapeutic proteins such as antibodies. However, proteins can attach to each other in millions of different combinations and the current docking solutions used to predict these interactions can be very slow. Faster and more accurate solutions are needed to streamline the process.

In a preprint Published earlier this year, a new machine learning model – EquiDock – was introduced that can quickly predict how two proteins interact. Unlike other approaches, the model does not rely on heavy candidate sampling and was found to make predictions up to 80–500 times faster than popular docking software.

To learn more about EquiDock and how artificial intelligence (AI) methods are moving the field of structural proteomics, Technology networks spoke with co-lead author of the article, Octavian-Eugen Ganeaa postdoctoral researcher in the MIT Computer Science and Artificial Intelligence Laboratory.

Molly Campbell (MC): For our readers who may not be familiar, can you describe your current research focus on proteomics?

Octavian Ganea (OG): My research uses AI (specifically deep learning) to model aspects of molecules that are important in various applications, such as drug discovery.

Proteins are involved in most biological processes in our body. Two or more proteins with different functions interact and form larger machines, ie complexes. They also bind to smaller molecules such as those found in drugs. These processes alter the biological functions of individual proteins. For example, an ideal drug would inhibit a cancer-causing protein by attaching to specific areas of the surface. I am interested in using deep learning to model these interactions and to help and accelerate the research of chemists and biologists by providing better and faster computational tools.

MC: How do AI-based methods advance the field of proteomics and specifically structural proteomics?

AND: Biological processes are very complex by nature and have their own mysteries, even for domain experts. For example, to understand how interacting proteins attach to each other, humans or computers need to try all possible attachment combinations to find the most plausible one. Intuitively, if you have two three-dimensional objects with very irregular surfaces, you have to rotate them and try to link them in every possible way until you can find two complementary regions on both surfaces that would fit very well in terms of their geometric and chemical patterns. This is a very time consuming process for both manual and computational approaches. In addition, biologists are interested in discovering new interactions between a very large number of proteins, such as the ~20,000 human proteome. This is important, for example, to automatically detect unexpected side effects of new treatments. Such a problem now becomes akin to an extremely large 3D puzzle where you have to simultaneously scan pieces to match, as well as understand how each single pairwise confirmation is done by trying all possible combinations and rotations.

MC: Can you explain how you made EquiDock?

AND: EquiDock takes the 3D structures of two proteins and instantly identifies which regions are likely to interact, which would otherwise be a complicated problem even for a biology expert. Discovering this information is then sufficient to understand how to rotate and orient the two proteins in their attached positions. EquiDock learns to capture complex coupling patterns from a large set of ~41,000 protein structures using a geometrically constrained model with thousands of parameters that are dynamically and automatically adjusted until they solve the task very well.

MC: What are the potential uses of EquiDock?

AND: As mentioned, EquiDock can enable rapid computerized scanning of drug side effects. This goes hand in hand with large-scale virtual screening of drugs and other types of molecules (eg antibodies, nanobodies, peptides). This is necessary to significantly shrink an astronomical search space that would otherwise be unfeasible for all of our current experimental capabilities (even aggregated globally). A rapid protein-protein docking method such as EquiDock combined with a rapid protein structure prediction model (such as AlphaFold2 developed by DeepMind) would aid in drug design, protein engineering, antibody generation or understanding the mechanism of action of a drug, among many other exciting applications critical in our quest for better disease treatments.

Octavian Ganea spoke with Molly Campbell, Senior Science Writer for Technology Networks.

Leave a Comment