Deep Learning for Validating and Estimating Resolution of Cryo-Electron Microscopy Density Maps †.

Todor Avramov,Javier Vargas,Swathi Adinarayanan,Josue Gomez-Blanco,Dan Vyenielo,Dong Si

doi:10.3390/molecules24061181

Todor Avramov, Javier Vargas + Show 4 more

Open Access

https://doi.org/10.3390/molecules24061181

Copy DOI

Abstract

Cryo-electron microscopy (cryo-EM) is becoming the imaging method of choice for determining protein structures. Many atomic structures have been resolved based on an exponentially growing number of published three-dimensional (3D) high resolution cryo-EM density maps. However, the resolution value claimed for the reconstructed 3D density map has been the topic of scientific debate for many years. The Fourier Shell Correlation (FSC) is the currently accepted cryo-EM resolution measure, but it can be subjective, manipulated, and has its own limitations. In this study, we first propose supervised deep learning methods to extract representative 3D features at high, medium and low resolutions from simulated protein density maps and build classification models that objectively validate resolutions of experimental 3D cryo-EM maps. Specifically, we build classification models based on dense artificial neural network (DNN) and 3D convolutional neural network (3D CNN) architectures. The trained models can classify a given 3D cryo-EM density map into one of three resolution levels: high, medium, low. The preliminary DNN and 3D CNN models achieved 92.73% accuracy and 99.75% accuracy on simulated test maps, respectively. Applying the DNN and 3D CNN models to thirty experimental cryo-EM maps achieved an agreement of 60.0% and 56.7%, respectively, with the author published resolution value of the density maps. We further augment these previous techniques and present preliminary results of a 3D U-Net model for local resolution classification. The model was trained to perform voxel-wise classification of 3D cryo-EM density maps into one of ten resolution classes, instead of a single global resolution value. The U-Net model achieved 88.3% and 94.7% accuracy when evaluated on experimental maps with local resolutions determined by MonoRes and ResMap methods, respectively. Our results suggest deep learning can potentially improve the resolution evaluation process of experimental cryo-EM maps.

Highlights

There are three principle methods used for determining the 3D atomic structures of proteins; they include X-ray crystallography, Nuclear Magnetic Resonance (NMR) spectroscopy, and 3D cryo-electron microscopy
The test accuracy of the trained 3D convolutional neural network (3D convolutional neural networks (CNNs)) model was assessed on a data set containing simulated maps that were not included in the training process and were never seen by the model
The 3D CNN model had higher prediction accuracy (99.75%) on the simulated test data set compared to the 92.73% of the dense artificial neural network (DNN) model

Summary

Introduction

There are three principle methods used for determining the 3D atomic structures of proteins; they include X-ray crystallography, Nuclear Magnetic Resonance (NMR) spectroscopy, and 3D cryo-electron microscopy (cryo-EM). The goal of these imaging techniques is to generate a high-quality and high-resolution, detailed protein macromolecule map that can be used in conjunction with other biochemical experiments and computational methods to generate a structural atomic model of the biomolecule. X-ray crystallography has been the gold-standard approach for solving protein structures for years [1]. Most protein structures with near-atomic or atomic resolutions have been determined by X-ray crystallography. Both techniques typically require large amounts of relatively pure samples (on the order of several milligrams)

Objectives

Methods

Results

Discussion

Conclusion