Protein Dynamics from Distance Predictions

Proteins are intrinsically dynamic and exist as an ensemble of conformations. Molecular dynamics simulations allow us to generate trajectories of proteins as they move around in a parameterized force field. These trajectories give us a deeper view of the conformations proteins take in solution than the static structures generated by experimental techniques such as X-ray crystallography. Many computational methods also exist for predicting three-dimensional structures for proteins with known amino acid sequences. Many of these methods use machine learning techniques to generate inter-residue distance predictions. Though the purpose of these distance predictions is to predict static structures that match experimentally determined protein structures, they may also contain information about the conformations the proteins adopt in solution. Analysis of protein trajectories and inter-residue distance predictions is helping us to better understand this relationship and may lead to the discovery of an efficient approach to determine dynamic characteristics from structure prediction data.