What is it about?
Approximately 40% of eukaryotic proteins do not adopt a rigid three-dimensional structures in cells, and instead are highly dynamic and disordered under physiological conditions. Many of these proteins fold into stable structures when they bind interaction partners in cells. The detailed molecular mechanisms by which disordered proteins fold-upon-binding their binding partners are relatively poorly understood at an atomic level. Here we develop a new computational method, employing deep learning, to analyze an atomic resolution computer simulations of a disordered protein folding-upon-binding its binding parter. Our method produces highly detailed atomic resolution insights into the step-wise process by which a disordered protein folds-upon-binding is physiological binding partner.
Featured Image
Photo by National Cancer Institute on Unsplash
Why is it important?
Analyzing simulations of disordered proteins is notoriously challenging. Disordered proteins have an extremely large number of degrees of freedom, making it difficult to extract meaningful mechanistic information for all-atom computer simulations. Our deep learning approach to build models of disordered protein dynamics and binding mechanisms provides a powerful new tool to analyze atomic resolution computer simulations - and will help produce mechanistic insight to interactions of disordered proteins involved in currently untreatable diseases.
Perspectives
A great deal of attention has been focused on application of deep learning approaching to computational chemistry and biophysics. As I have watched this development - I have been curious to what extent deep learning is enabling us to truly discover new things as opposed to learning to rediscover things we essentially already knew more efficiently. When we started this project - I wasn't focused on using deep learning methods - and when graduate student Tommy Sisk began to experiment with applying neural networks to build mechanistic models of protein dynamics - I was truly to curious to see how they cold improve upon the existing state-of-art in terms of building atomic resolution models of protein dynamics. Ultimately I was pleasantly surprised to find that the deep learning approach we developed enabled us to discover many new things about the system we study. I am excited to continue to develop these deep learning approaches for analyzing computer simulations of dynamic proteins in the future.
Paul Robustelli
Dartmouth College
Read the Original
This page is a summary of: Folding-upon-binding pathways of an intrinsically disordered protein from a deep Markov state model, Proceedings of the National Academy of Sciences, January 2024, Proceedings of the National Academy of Sciences,
DOI: 10.1073/pnas.2313360121.
You can read the full text:
Contributors
The following have contributed to this page
