AI, molecular simulations get to the root of better plants

By combining AI with molecular dynamics simulations, researchers at the Department of Energy’s Oak Ridge National Laboratory have developed a new tool to more accurately predict how plants and helpful microbes communicate and form partnerships at the most fundamental level.

The new AI-powered workflow helps scientists identify which plant genes control the best microbial partnerships. This accelerates the engineering of microbiomes that help plants grow faster, require less fertilizer, and produce more biomass for conversion into valuable fuels, chemicals and materials. The approach speeds up research for the nation’s energy and food security and increases U.S. competitiveness in the global biotechnology sector.

Plants and the microbes in their environment communicate using chemical signals called ligands to form partnerships that boost plant growth and health. A key class of ligands, lipo-chitooligosaccharides (LCOs), are of interest to ORNL scientists since they facilitate this plant-microbe symbiosis. However, predicting which proteins — large, complex molecules in cells that carry out specific tasks — will recognize and precisely bind with these signals has been challenging due to the flexible and large molecular structure of the LCOs.

Existing computational tools such as the AlphaFold program that predicts the three-dimensional shapes of proteins offer limited help because they have been mainly trained on datasets made up of smaller, drug-like ligands. AlphaFold also typically predicts static interactions rather than the dynamic fluctuations at play in LCOs. 

To generate better protein-ligand matching predictions, ORNL biophysicists developed a hybrid approach combining molecular dynamics (MD) simulation, which allows for broader sampling of protein dynamic structures, with machine learning (ML) prediction. The ML models were trained on a large dataset of protein-ligand complexes. 

The simulations were performed on two of the nation’s fastest supercomputers, Frontier and Summit, at the Oak Ridge Leadership Computing Facility, a DOE Office of Science user facility at ORNL dedicated to open science to accelerate U.S. innovation and competitiveness.

The method, dubbed MD/ML, ranks how strongly plant receptors bind to the ligands, and it works even when starting protein structures are mere rough models. The research predicted binding that matched experimental lab results and revealed new structural details about how the binding happens.

Molecular matchmaking accelerates plant transformation

“Being able to quickly predict these molecular ‘match-ups’ means scientists can focus their experiments — saving time and money,” said Erica Prates, project co-lead in the ORNL Computational and Predictive Biology Group. 

“The technique showed that we could predict the relative strength of large, highly flexible ligands to protein receptors,” said Omar Demerdash, project co-lead in ORNL’s Molecular Biophysics Group. “That binding strength ultimately governs what happens inside cells, including which genes are switched on and a myriad of other physiological processes. This is key to understanding how plants interact with microbes, or how medicines work in the human body to treat disease.”

“This method takes into account the reality that proteins aren’t rigid — they’re wiggling all the time,” said Dan Jacobson, a corresponding author on the paper and a computational systems biologist at ORNL. “But most of our protein structure prediction tools to date end up generating a static structure, resulting in a rigid view disconnected from real-life flexibility. By performing molecular simulations that take into account protein motion, you get a much better way of finding these binding events.” 

“We’ve developed a way to get a much better understanding of the interface of these receptors in plants and the external microbial world,” Jacobson said. “It’s a nice addition to our toolkit for plant-microbe interactions research, and has broader applications such as exploring the repurposing of existing drug therapies to treat health disorders.”  

Other ORNL scientists on the team include Tomás Rush, Udaya Kalluri and Manesh Shah. The research was supported by the Plant-Microbe Interfaces Science Focus Area, part of the DOE Office of Science Biological and Environmental Research program. 

UT-Battelle manages ORNL for DOE’s Office of Science. The single largest supporter of basic research in the physical sciences in the United States, the Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit energy.gov/science.