Nanobiotics: AI discovering where and how nanoparticles bind to proteins

Applications of AI

The protein appears as a lumpy gray mass made up of hundreds of balls stuck together. One of the long clumps has a roughly U-shaped nanoparticle consisting of about six balls wrapped around it. The protrusion is about 2-3 times the diameter of a single ball in length and 1-2 times in diameter.
A new computer model, NeCLAS, predicts that nanoparticles, shown as a series of yellow balls attached with a web, fit neatly around very specific protrusions on the protein marked in blue. The binding site is confirmed experimentally. This class of nanoparticles, called molecular tweezers, can be used to get into the action of pathogens and aggregates of toxic proteins. Image credit: Paolo Elvati, Violi Lab, University of Michigan.

Determining whether and how nanoparticles and proteins bind to each other is an important step toward being able to design antibiotics and antivirals on demand, and is being studied at the University of Michigan. A developed computer model makes it possible.

This new tool could help find ways to thwart antibiotic-resistant infections and emerging viruses, as well as design nanoparticles for a variety of purposes.

“In 2019 alone, 4.95 million people died from antibiotic resistance. By 2050, antibiotic resistance will kill 10 million people, even before COVID-19 exacerbated the problem.” The study shows that the Cover of Nature Computational Science.

“In my ideal scenario, 20 or 30 years from now, I would like to be able to rapidly produce the best nanoparticles that can treat any superbug.”

Much of the work in cells is done by proteins. Surface interaction sites stitch and tear apart molecules, open doorways to cells, break down sugars to release energy, build structures to support cell clusters, and more. can be modified. If we could design drugs that target key proteins in bacteria and viruses without harming our own cells, humanity could rapidly fight emerging diseases.

Dubbed NeCLAS, the new model uses machine learning, an AI technology that powers the virtual assistant and ChatGPT on smartphones. But instead of learning to process language, it absorbs structural models of proteins and their known interaction sites. From this information, infer how proteins and nanoparticles interact, predict binding sites and potential binding between them, and predict interactions between two proteins or two nanoparticles. learn to

“Other models exist, but ours is the best for predicting the interactions between proteins and nanoparticles,” said Paolo Elvati, UM Associate Fellow in Mechanical Engineering.

For example, AlphaFold is a widely used tool for predicting the 3D structure of proteins based on building blocks called amino acids. This ability is important, but it’s only the beginning. Discovering how these proteins assemble to form larger structures and designing practical nanoscale systems is the next step.

“That’s where NeCLAS comes in,” says Jacob Saldinger, a Ph.D. student in chemical engineering at UM and lead author of the study. “By showing how nanostructures interact, we go beyond AlphaFold and are not limited to proteins. can be transformed.”

The team tested three case studies that provided additional data.

  • molecular tweezers. A molecule binds to a specific site on another molecule. This approach can stop harmful biological processes such as protein plaque aggregation in brain diseases like Alzheimer’s disease.
  • How graphene quantum dots disrupt biofilms produced by staphylococci. These nanoparticles are flakes of carbon, only a few atomic layers thick and 0.0001 millimeters on a side. Disrupting biofilms may be an important tool in combating antibiotic-resistant infections such as methicillin-resistant Staphylococcus aureus (MRSA), a superbug common in hospitals.
  • Whether graphene quantum dots disperse in water. This indicates that the model can predict the binding between nanoparticles even though it was trained on protein-protein data only.

Many protein-protein models set the amino acid as the smallest unit that the model should consider, but this does not work for nanoparticles. Instead, the team set the size of its smallest features to be roughly the size of an amino acid, but then let the computer model decide where the boundaries between these smallest features were. The result is a representation of proteins and nanoparticles resembling ensembles of interconnected beads, allowing for more flexible exploration of small-scale interactions.

“Besides being more versatile, NeCLAS uses much less training data than AlphaFold. We only have 21 nanoparticles to investigate, so we need to use the protein data wisely,” he says. said Matt Raymond, Ph.D. student in electrical and computer engineering and study co-author.

Next, the team plans to investigate other biofilms and microorganisms, including viruses.

Research in Nature Computational Science was funded by the University of Michigan Blue Sky Initiative, the Army Research Service, and the National Science Foundation.

Violi is also a professor of electrical engineering, computer engineering, chemical engineering and biophysics.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *