• Tue. Feb 27th, 2024

An efficient machine learning pipeline predicts the location of nanoscale interactions

An efficient machine learning pipeline predicts the location of nanoscale interactions

Recognizing how a nanoparticle and a protein bind to each other is an important step toward designing antibiotics and antivirals as needed, and a computer model developed at the University of Michigan can do just that.

The new tool could help find ways to block antibiotic-resistant infections and new viruses—helping in the design of nanoparticles for a variety of purposes.

“In 2019, the number of deaths due to antimicrobial resistance was 4.95 million. Even before the Covid-19 crisis worsens, studies show that by 2050 the number of deaths due to antibiotic resistance will reach 10 million,” said Arthur F. said Angela Violi, Thurnau professor of mechanical engineering and corresponding author of the study. Cover of Nature Computational Science.

In my ideal scenario, 20 or 30 years from now, I want to rapidly produce the best nanoparticles that can treat any superbug.


Angela Wylie, Arthur F. Thurnau Professor, Mechanical Engineering, University of Michigan

Proteins do most of the work inside cells. Interaction sites on their surfaces can join molecules together, pull them apart, and make other changes — opening doors to cells, breaking down sugars to release energy, building structures to support groups of cells, and more. If we can design drugs that target critical proteins in bacteria and viruses without harming our own cells, it could enable humans to quickly fight new and changing diseases.

The new model, called NeCLAS, uses machine learning—an AI technology that powers the virtual assistant on your smartphone and ChatGPT. But instead of learning to process language, it absorbs structural models of proteins and their known interaction sites. From this information, it learns to predict how the proteins and nanoparticles will interact, the binding sites and the probability of binding between them—as well as predicting interactions between two proteins or two nanoparticles.

“Other models exist, but ours is the best for predicting interactions between proteins and nanoparticles,” said Paolo Elvati, UM associate research scientist in mechanical engineering.

For example, AlphaFold is a widely used tool for predicting the 3D structure of a protein based on its building blocks, called amino acids. While this capability is critical, it is only the beginning: figuring out how these proteins assemble into larger structures and designing practical nanoscale systems are the next steps.

“That’s where NeCLAS comes in,” said Jacob Saldinger, a UM doctoral student in chemical engineering and first author of the study. “It goes beyond alphafold by showing how nanostructures can interact with each other, and it’s not just limited to proteins. This enables researchers to understand potential applications of nanoparticles and optimize their designs.

The team tested three case studies for which they have additional data:

  • Molecular tweezers, in which one molecule binds to a specific site on another molecule. This approach could prevent harmful biological processes such as the aggregation of protein plaques in brain diseases such as Alzheimer’s.
  • How graphene quantum dots disrupt the biofilm produced by staph bacteria. These nanoparticles are flakes of carbon, no more than a few atomic layers thick and 0.0001 millimeters on a side. Disrupting biofilms is a critical tool for fighting antibiotic-resistant infections — including the superbug methicillin-resistant Staphylococcus aureus (MRSA), commonly acquired from hospitals.
  • Graphene quantum dots will disperse in water, demonstrating the model’s ability to predict nanoparticle-nanoparticle binding despite being trained exclusively on protein-protein data.

While many protein-protein models set amino acids as the smallest unit to be considered by the model, this does not work for nanoparticles. Instead, the team set the size of that smallest feature to roughly the size of an amino acid, but let the computer model decide where the boundaries between these smallest features lie. The result is representations of proteins and nanoparticles that look like collections of interconnected beads, providing greater flexibility in exploring small-scale interactions.

“Besides being more general, NeCLAS also uses less training data than Alphafold. We only had 21 nanoparticles to look at, so the protein data had to be used smartly,” said study co-author Matt Raymond, a UM doctoral student in electrical and computer engineering.

Next, the team plans to explore other biofilms and microbes, including viruses.

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

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

Source:

Journal Reference:

Saldinger, JC, etc. (2023). Domain-agnostic predictions of nanoscale interactions in proteins and nanoparticles. Nature Computational Science. doi.org/10.1038/s43588-023-00438-x.

Leave a Reply

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