AI System ‘FragFold’ Predicts Protein Fragments for Targeted Binding and Inhibition
In a significant stride for both basic research and potential therapeutic applications, scientists at MIT have developed a novel artificial intelligence (AI) system named FragFold that can predict protein fragments capable of binding to and inhibiting target proteins. This innovative tool, detailed in a study published in the Proceedings of the National Academy of Sciences, builds upon existing AI models to offer a powerful means of understanding and manipulating protein interactions [1].
Proteins are the workhorses of biological functions, and how they interact with each other dictates everything from DNA transcription to cell division. Understanding these interactions at a molecular level remains a challenge, but short protein fragments have emerged as functional entities capable of altering a protein’s function by binding to its interfaces.
FragFold leverages AlphaFold, an AI model known for its protein folding and interaction prediction capabilities. The researchers experimentally validated that over half of FragFold’s predictions for binding or inhibition were accurate, even without prior structural data on interaction mechanisms. This has profound implications for studying proteins with unknown functions or structures.
“Our results suggest that this is a generalizable approach to find binding modes that are likely to inhibit protein function, including for novel protein targets,” explains Andrew Savinov, co-first and corresponding author and a postdoc in the Li Lab. He emphasizes the tool’s applicability to proteins lacking known functions or interactions, stating, “We can put some credence in these models we’re developing.”
One notable application of FragFold was in studying FtsZ, a key protein for cell division, particularly its intrinsically disordered region that has been challenging to study with traditional methods. By exploring fragments of FtsZ, FragFold identified new binding interactions, confirming and expanding upon previous experimental measurements of FtsZ’s biological activity.
Amy Keating, Jay A. Stein (1968) Professor of Biology, professor of biological engineering, and department head, highlights the broader impact of this work: “This is one example of how AlphaFold is fundamentally changing how we can study molecular and cell biology. Creative applications of AI methods, such as our work on FragFold, open up unexpected capabilities and new research directions.”
The FragFold system involves computationally fragmenting proteins and modeling how these fragments bind to interaction partners. High-throughput experimental measurements in millions of cells, each producing a protein fragment, were then used to compare predicted binding maps to actual effects in living cells.
Researchers also explored the complex between lipopolysaccharide transport proteins LptF and LptG, discovering that a protein fragment of LptG inhibited this interaction, disrupting the delivery of lipopolysaccharide, a crucial component of the E. coli outer cell membrane. This confirmed FragFold’s capability to accurately predict binding that leads to inhibition.
Looking ahead, Savinov is interested in exploring fragment functions beyond inhibition, such as protein stabilization, function alteration, or triggering protein degradation. Gene-Wei Li, associate professor of biology and Howard Hughes Medical Institute investigator, envisions using functionalized fragments to modify native proteins, change their subcellular localization, and even reprogram them to create new tools for studying cell biology and treating diseases.
