Machine learning to estimate the local quality of protein crystal structures

Ikuko Miyaguchi,Mitsunori Ikeguchi,Hiroyuki Nakagawa,Miwa Sato,Yuichi Kokabu,Akiko Kashima,Atsushi Tokuhisa,Biao Ma,Shigeyuki Matsumoto,Masateru Ohta

doi:10.1038/s41598-021-02948-y

Abstract

Low-resolution electron density maps can pose a major obstacle in the determination and use of protein structures. Herein, we describe a novel method, called quality assessment based on an electron density map (QAEmap), which evaluates local protein structures determined by X-ray crystallography and could be applied to correct structural errors using low-resolution maps. QAEmap uses a three-dimensional deep convolutional neural network with electron density maps and their corresponding coordinates as input and predicts the correlation between the local structure and putative high-resolution experimental electron density map. This correlation could be used as a metric to modify the structure. Further, we propose that this method may be applied to evaluate ligand binding, which can be difficult to determine at low resolution.

Highlights

Low-resolution electron density maps can pose a major obstacle in the determination and use of protein structures
An electron density map for X-ray crystallography or a Coulomb potential map for cryogenic electron microscopy (cryo-EM) is calculated from experimental data, and a protein structure is constructed by placing the atoms of the protein according to the map
We selected 22 proteins from them based on conditions, such as whether there are homologous protein structures that could be used as homology modeling templates and whether they belonged to different categories in the CATH classification and so on

Summary

Introduction

Low-resolution electron density maps can pose a major obstacle in the determination and use of protein structures. We describe a novel method, called quality assessment based on an electron density map (QAEmap), which evaluates local protein structures determined by X-ray crystallography and could be applied to correct structural errors using low-resolution maps. QAEmap uses a three-dimensional deep convolutional neural network with electron density maps and their corresponding coordinates as input and predicts the correlation between the local structure and putative high-resolution experimental electron density map. An electron density map for X-ray crystallography or a Coulomb potential map for cryo-EM is calculated from experimental data, and a protein structure is constructed by placing the atoms of the protein according to the map. While the secondary structures and the rigid part of the inner molecule provide clear electron density maps, the surrounding or loop regions are often obscured; the side chains are often more obscured than the main chain This is because of the thermal vibration of atoms, multiple conformations, or regional disorder. In all these cases, the experimenter needs to make subjective decisions

Methods

Results

Conclusion