With AI-powered prediction, Spartan scientists target potential biothreats
Researchers at Michigan State University are spearheading a new effort to transform how the world identifies and responds to emerging biological dangers.
Composed of experts in biophysics, molecular modeling and AI, the group – known as “Team Green” – is part of a new national scientific moonshot supported by the Defense Advanced Research Projects Agency, or DARPA.
If successful, the Spartans would help create an AI tool that almost instantly identifies a protein’s function by analyzing its complex movements. This tool would potentially reshape our understanding of basic biology, as well drastically shrink the time needed to identify and counter unknown biothreats.
“Imagine being able to detect new molecules in the environment and know immediately whether they are a threat, and what effects they would have in the body,” said Alex Dickson, team member and professor in MSU’s Department of Biochemistry & Molecular Biology, or BMB.
“You almost picture something from Star Trek.”
Tight timelines
From antibodies and enzymes to hormones and hemoglobin, life runs on proteins. In our bodies alone, you’ll find almost 20,000 different types, each with a unique shape that dictates its role.
Using AI tools such as Google DeepMind’s AlphaFold — which received a portion of the 2023 Nobel Prize in Chemistry — scientists can now predict a protein’s 3D structure faster and more accurately than ever before.
Knowing what a protein looks like, however, is only half the battle.
Proteins are lively collections of molecules that constantly vibrate, twist and contort, with different shapes leading to specific biochemical possibilities. If researchers want to know how a new drug or an unknown pathogen actually functions, they need to know how its proteins act in real time.
To see a fuller picture, researchers will simulate "molecular movies.” Though powerful, this process is highly resource intensive. Predicting a single microsecond of protein behavior can take up to a week of heavy computing that models millions of atoms and billions of microscopic events.
This is where the Network of Optimal Dynamic Energy Signatures, or NODES program, steps in. Launched by DARPA in 2025, NODES is shifting the scientific focus to how a protein moves — its dynamics — as the best way to determine its biological behavior with lightning speed.
With $1.1 million in funding, MSU’s Team Green is just one of a handful of groups across the nation selected for the initiative.
A defense-oriented agency, DARPA sees NODES as a necessary breakthrough to stay ahead of both natural and human-made biothreats, and most of all reduce the time it takes to respond to these dangers, going from months and weeks to hours and days.
“Let’s say someone creates a designer protein with nefarious properties” said Josh Vermaas, an assistant professor in BMB and MSU’s Plant Research Laboratory.
“We need to understand the dynamics of that protein so we know what those properties are — and we need to know very quickly.”
Alongside Vermaas and Dickson, the collaboration includes researchers Michael Feig, Guowei Wei, and Daniel Woldring, all leading figures in biology, mathematics and molecular modeling.
Together, they’ll use supercomputers to generate massive quantities of synthetic data about protein movements, which they’ll then use to train an AI model in an ever-improving feedback loop.
By constantly refining itself, the model would train for the day it’s asked to crack the code of a real-world mystery protein, ensuring the next global response happens in hours rather than years.
When working, the AI-powered tool would also fundamentally improve the pace of protein science, saving researchers time, resources and paving the way for untold breakthroughs.
Any opinions, findings, conclusions, or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Defense Advanced Research Projects Agency (DARPA).
- Categories: