“Protein structure modeling and assessment using deep learning (DeepMainmast and DAQ score)”
_
Date:
4:00 PM – 5:30 PM Wednesday, October 23th, 2024,
Location: Hybrid (CryoEM-Lab KEK and Zoom)
Speaker
Tsukasa Nakamura, PhD
Structural Biology Research Center,
Institute of Materials Structure Science,
High Energy Accelerator Research Organization (KEK), Japan
formerly: Biological Sciences, Purdue University, USA
Profile
2019 Ph.D. (Science) in Bioinformatics, The University of Tokyo, Japan
2019-2022 Postdoctoral Fellow (JSPS(PD)), Tohoku University, Japan
2022-2024.6 Postdoctoral Research Associate, Purdue University, USA
2024.7-current Postdoctoral Research Fellow, KEK IMSS SBRC, Japan
Abstract
Cryogenic electron microscopy (cryo-EM) has become one of the main experimental methods for determining protein structures. In cryo-EM, protein structure modeling is in general more difficult than X-ray crystallography as the resolution of maps is often not high enough to specify atom positions. We have been developing computational methods for modeling protein structures from cryo-EM maps determined at a medium resolution (2.5-5 Å). For maps at medium resolution, it turned out that deep learning can provide useful structure information for modeling and protein model quality assessment. In this seminar, we present deep learning applications in protein structure modeling (DeepMainmast) and protein model quality assessment (DAQ score) along with the large-scale analysis results of the quality of protein models from cryo-EM in the Protein Data Bank, which is accessible at DAQ-Score Database https://daqdb.kiharalab.org/. All the tools we developed are easy to use at https://em.kiharalab.org/.