Protein dynamics: bringing together simulations and experiments to understand biology at the atomic scale

Structural biology has undergone revolutionary developments over the last few years: on the experimental side, cryo-electron microscopy has made a leap forward and has established itself as a very powerful method to obtain structures of biomolecules, from large complexes to ever smaller and more challenging targets. Together with experimental structures from X-ray crystallography and nuclear magnetic resonance, nearly 200,000 protein structures have been solved to date. On the other hand, and based on those structures together with machine learning, the computational leap in structure prediction using AlphaFold2, RosettaFold and other computational tools, has made it possible for anyone to obtain quite reliable models of protein structures from not more than a sequence. One might argue that structural biology is a solved problem.

Yet, the enormous amount of details contained in these structures is somewhat deceptive, as they represent a single static snapshot obtained, in most cases, by trapping a protein inside a frozen crystal or a thin layer of vitreous ice. Biology, however, is dynamic, almost by its very definition. Processes such as enzymatic turnover, protein transport and folding via biomolecular machinery (chaperoning), or the translocation of small molecules across biological membranes fundamentally involve the sampling of multiple conformations of the proteins involved, i.e., dynamics. Moreover, the realization that many proteins are functional despite lacking any defined 3D structure, i.e. so-called intrinsically disordered proteins, has severely challenged our view of the structure-function relationship. Thus, in order to understand biological processes, it is instrumental to decipher the motions of proteins.

The study of complex biomolecular dynamics is a considerable challenge, as it requires not only to determine a set of atomic coordinates in 3D space, but also how they evolve in a fourth dimension, time. No single experimental technique is able to achieve this task by itself. Similarly to the methodological breakthroughs described above for structure determination, the methodology for characterizing protein dynamics has been truly remarkable, too. On the experimental side, advances for studying protein dynamics include revolutionary new facilities, such as X-ray free electron lasers (XFEL), which are extremely bright, 10¹⁰ times more than synchrotrons, enabling completely new ultra-fast kinetics experiments. Ingenious other developments include ultra-fast atomic-force microscopes, or ultra-fast mixing devices for trapping proteins in different states for electron microscopy. Often these developments inspire each other, and, for example, developments for XFELs are finding applications in EM. The integration of multiple experimental methods nowadays allows, e.g. to understand intrinsically disordered proteins with unprecedented detail, using single-molecule and bulk methods.

The progress of the experimental characterization of protein dynamics greatly benefits from computational methods. Continuous developments in computational approaches and theory have significantly pushed the frontiers of computational protein dynamics studies. In particular, new enhanced sampling approaches, often augmented by machine learning, have expanded the spatial and temporal domains accessible to state-of-the-art computational models, and has made them commensurate to time scales accessed experimentally. This situation provides excellent opportunities for joint computational and experimental protein dynamics studies. Much progress has been achieved by combining data from different experimental techniques, and often such integration is achieved through molecular simulations. Connecting experimental with computational data has turned out to be game-changing in order to understand the function of increasingly complex biomolecular systems at ever-growing details.

Important challenges remain to be overcome both for experimental and computational approaches. For example, a particularly important and active research field in computational biophysics aims at simulating molecular dynamics over sufficient time scales (micro- to milliseconds) to observe the biologically relevant conformational changes, which coincide with many biomolecular processes (e.g. enzymatic catalysis). Computational investigation of millisecond events in large biological objects is computationally very expensive. Enhanced sampling and accelerated simulation methods are often limited by our ability to identify the relevant degrees of freedom at play and sample them with utmost efficiency. Pushing the limits of computer simulations with new hardware and algorithms to longer time scales and larger system sizes is therefore a topic of central importance. Another active research area is related to understanding the solvation of biomacromolecules through computational and experimental approaches.

It is of great importance for this entire (fairly heterogeneous) community of researchers on protein dynamics, to meet, exchange ideas, challenge each other and think together about the next big challenges for the field. This is the ambition of the Les Houches-Telluride Workshop on Protein Dynamics. It brings together the latest developments in various computational and experimental fields, aiming to stimulate mutual benefits from this cross-disciplinary contact.