The AI That Can Design Complex Protein Structures: Expanding the Frontiers of Protein Research

Monday - March 11 2024, 17:18 UTC - 1 year ago

Proteins are dynamic and social creatures, and AI has become a powerful tool in understanding their interactions and designing new therapies. The latest AI algorithm, RoseTTAFold All-Atom, incorporates a variety of biomolecules, creating complex structures with potential therapeutic applications. This is a major advancement in protein research and could lead to even more breakthroughs in the future.

Proteins are not just static machines, but dynamic creatures that are constantly transforming to fulfill a cell's needs. They also didn't exist in isolation, but interact with many other biomolecules in a complex dance. These interactions are the foundation of many biological processes, from gene regulation to cell signaling to maintaining cognitive function. Harnessing these connections has led to numerous therapies and treatments, and further advances could be accelerated by the power of AI .

In February 2024, a study published in Science introduced us to the newest tool in the AI-protein research toolbox: RoseTTAFold All-Atom. Led by Dr. David Baker at the University of Washington, this AI builds on the previous RoseTTAFold algorithm, which only focused on proteins, to incorporate a wide variety of other biomolecules. This includes DNA, RNA, and small molecules like iron, which are crucial for many protein functions .

What sets RoseTTAFold All-Atom apart is its ability to learn and understand not just proteins, but a range of other biomolecules, without any prior knowledge of their 3D structure. This means that it can map out complex molecular machines and interactions at an atomic level, giving scientists unprecedented insight into the inner workings of cells. But this isn't just a theoretical exercise. In the study, RoseTTAFold All-Atom was paired with generative AI to create proteins that easily grabbed onto a heart disease medication and regulate important molecules like heme and bilin .

These examples are just the beginning; the team is making the algorithm available to the public so that scientists can create even more complex structures and potentially discover new therapies. A decade ago, we would have never thought that AI could dream up new proteins, but now it is a reality. In 2020, Google DeepMind's AlphaFold and Baker Lab's RoseTTAFold solved a problem that had puzzled scientists for half a century - predicting protein structures .

This breakthrough opened the floodgates for generative AI, which has since been used to create a range of designer proteins with diverse functions, including regulating hormones, acting as artificial enzymes, and changing shape like transistors in electronic circuits. But here's the problem: these AI models have tunnel vision. They are too precise and focused on proteins, missing out on potential therapeutic opportunities by not looking at the bigger picture .

That's where RoseTTAFold All-Atom comes in. By expanding the scope of AI to incorporate a wide range of biomolecules, it has the potential to unlock a whole new world of designer proteins and open up new avenues for treating diseases and improving human health.