Deep Learning Protein Conformational Space with Convolutions and Latent Interpolations

Ramaswamy, Venkata K.; Musson, Samuel C.; Willcocks, Chris G.; Degiacomi, Matteo T.

Abstract

Determining the different conformational states of a protein and the transition paths between them is key to fully understanding the relationship between biomolecular structure and function. This can be accomplished by sampling protein conformational space with molecular simulation methodologies. Despite advances in computing hardware and sampling techniques, simulations always yield a discretized representation of this space, with transition states undersampled proportionally to their associated energy barrier. We present a convolutional neural network that learns a continuous conformational space representation from example structures, and loss functions that ensure intermediates between examples are physically plausible. We show that this network, trained with simulations of distinct protein states, can correctly predict a biologically relevant transition path, without any example on the path provided. We also show we can transfer features learned from one protein to others, which results in superior performances, and requires a surprisingly small number of training examples.

Citation

Ramaswamy, V. K., Musson, S. C., Willcocks, C. G., & Degiacomi, M. T. (2021). Deep Learning Protein Conformational Space with Convolutions and Latent Interpolations. Physical Review X, 11(1), Article 011052. https://doi.org/10.1103/physrevx.11.011052

• Journal Article Type: Article
• Acceptance Date: Jan 26, 2021
• Online Publication Date: Mar 15, 2021
• Publication Date: 2021-03
• Deposit Date: Mar 20, 2021
• Publicly Available Date: Oct 22, 2021
• Journal: Physical Review X
• Electronic ISSN: 2160-3308
• Publisher: American Physical Society
• Peer Reviewed: Peer Reviewed
• Volume: 11
• Issue: 1
• Article Number: 011052
• DOI: https://doi.org/10.1103/physrevx.11.011052
• Public URL: https://durham-repository.worktribe.com/output/1245217