As artificial intelligence (AI) continues to evolve rapidly, it’s interesting to regularly ask what AI means for Folding@home.
In a nutshell, the current state of AI opens up lots of opportunities for Folding@home! Below are a few examples.
Structure prediction vs dynamics: With the advent of tools like AlphaFold, I think it is fair to say that the structure prediction problem is largely solved. Questions related to dynamics, like how do proteins fold and what role do structural dynamics play in function, remain open. These questions are the main focus of research with Folding@home, so we still have a lot of work ahead of us. Fortunately, improved methods for structure prediction is helping accelerate our work, e.g. by giving us starting points for simulations of proteins that don’t have experimental structures yet. Our data is also providing a foundation for efforts to predict protein dynamics with AI.
Data analysis: AI is a really powerful way to find patterns in complex data. We’re leveraging this feature all the time! For example, we’re using AI to compare the dynamics of properly functioning proteins vs those implicated in disease, identify distinguishing features, and inform the development of new therapeutics.
Software development: It takes a lot of code to make Folding@home run smoothly. As with many software projects, we are benefiting greatly from using AI to generate and or debug code.