Extraction of Protein Dynamics Information Hidden in Cryo-EM Map Using Deep Learning [article]

Shigeyuki Matsumoto, Shoichi Ishida, Mitsugu Araki, Takayuki Kato, Kei Terayama, Yasushi Okuno
2020 bioRxiv   pre-print
Protein functions are associated with their three-dimensional structures and in-solution dynamics. After the technical breakthroughs in cryogenic electron microscopy (cryo-EM) single-particle analysis (SPA), numerous structures have been solved at atomic ~ near-atomic resolutions, including extremely large macromolecules that were not solved by conventional techniques. Their dynamics analysis based on the solved structures further deepens the understandings of the functional mechanisms.
more » ... the elucidations are often hampered by the large molecular sizes and the complicated structural assemblies making both the experimental and computational approaches challenging. Here, we report a deep learning-based approach, DEFMap, to extract the dynamics information "hidden" in a given cryo-EM density map, using a three-dimensional convolutional neural network (3D-CNN). DEFMap successfully provided dynamics information equivalent to molecular dynamics (MD) simulation and experimental approaches only from cryo-EM maps. Indeed, DEFMap has detected dynamics changes associated with molecular recognitions and the accompanying allosteric conformational stabilizations. We expect that this new approach allows cryo-EM SPA to quantitatively grasp the in-solution protein behavior at atomic and residue levels beyond the static information, gaining biological insights into the functional mechanisms.
doi:10.1101/2020.02.17.951863 fatcat:f5vbnt6jw5c33lrxtnbognrw5m