This Artificial Intelligence Paper Introduces AF2Complex: A Deep Learning Tool Designed To Predict The Physical Interactions of Multiple Proteins

Large, complex molecules called proteins are in charge of nearly all of the vital processes that occur within the bodies of living organisms. Because of this, although being a relatively recent field of study, protein research and engineering have emerged as fundamental. One major breakthrough in protein research was the introduction of AlphaFold and AlphaFold 2 by DeepMind, a subsidiary of Alphabet. AlphaFold is a machine learning tool that can precisely predict the three-dimensional structures of proteins. Nevertheless, despite these advancements, it is extremely challenging and time-consuming to analyze the folding and transport of biological proteins experimentally.

DeepMind researchers have once again demonstrated that their most recent model, AF2Complex, can offer assistance in resolving this problem. AlphaFold 2 Complex, or simply AF2Complex, is a deep learning technique created to predict the physical interactions between several proteins. In an unprecedented level of detail, AF2Complex can predict which proteins will interact with one another to form functional complexes. The pace of protein research is being substantially accelerated by this model, which is heavily based on DeepMind’s AlphaFold 2, a machine learning tool that can predict the three-dimensional structures of proteins using only their amino acids. DeepMind’s work has also gained recognition in the esteemed biomedical scientific journal eLife.

Using amino acids, AF2Complex can predict whether proteins can interact to form functional complexes, which parts of each structure are most likely to interact, and even which protein complexes are most likely to combine to generate supercomplexes. Supercomplexes are enormous, interdependent protein clusters required for biological functions. The successful creation of AF2Complex led to the belief that this method has vast potential for locating and describing the collection of protein-protein interactions crucial to life. Researchers studying proteins mostly conduct computational studies to understand the atomic features of supercomplexes. AF2Complex, which acts as a “computational microscope” supported by deep learning and supercomputing, is essential to this research since it has significantly enhanced the speed of the investigation. 

Another research area focuses on clarifying the interactions between the proteins in the pathway that transports a newly generated protein from the inner to the outer membrane of the bacteria. This complex protein synthesis and transport pathway process are also being studied using AF2Complex. Discoveries during this research may provide new targets for developing antibiotics and treatments and a foundation for leveraging AF2Complex to speed up this type of biomedical research computationally in general.

In order to persuade the molecular biology community of the tool’s potency and high impact, the researchers decided to apply AF2Complex to a pathway in Escherichia coli (E. coli), a model organism frequently used for experimental DNA manipulation and protein synthesis due to its relative simplicity and rapid growth. The team looked at the production and movement of outer membrane proteins (OMPs), which are found on gram-negative bacteria like E. coli’s outermost membrane and are essential for nutrition exchange. Since proteins are created inside the cell, they need to be moved to reach the outer membrane.

The researchers conducted several experimental evaluations to determine which protein pairs the tool predicted to interact and which pairs were likely to form supercomplexes. For this purpose, they compared a few proteins crucial for synthesizing and transporting OMPs to roughly 1,500 other proteins (all known proteins in E. coli’s cell envelope). The researchers matched the tool’s predictions to previously published experimental data to validate their findings. Most of the formerly known interacting pairs, and several unknown ones, were accurately predicted by AF2Complex. The tool’s accuracy was also strengthened by its ability to draw attention to structural aspects of those interactions that explain data from earlier tests.

The researchers are optimistic that AF2Complex could significantly impact biomedical research. The key proteins studied by the tool in the pathway could serve as brand-new targets for developing new antibiotics. Predicting a supercomplex’s structural model is extremely complicated in contrast to predicting the structures of a single protein sequence. In this sense, AF2Complex might be a new computational tool for biologists to perform experiments using various protein combinations, accelerating the pace and efficiency of this particular biological research field.

Check out the Paper. All Credit For This Research Goes To the Researchers on This Project. Also, don’t forget to join our Reddit Page, Discord Channel, and Email Newsletter, where we share the latest AI research news, cool AI projects, and more.