Molecular dynamic (MD) simulations have been actively used in the study of protein structure and function. However, extensive sampling in the protein conformational space requires large computational resources and takes a prohibitive amount of time. In this study, we used the enzyme adenosine kinase (ADK) as an example and applied variational autoencoders (VAEs), a deep learning model, to explore the protein conformational space through MD simulations. VAEs are shown to be superior to traditional autoencoders (AEs) by a benchmark study, with low deviation between the training and decoded conformations. Moreover, we showed that the learned latent space can be used to generate unsampled protein conformations. Additional simulations starting from these generated conformations can capture biological-relevant transitions among conformational states with significantly less sampling.

Exploration of protein conformational space with variational autoencoder

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.