Computational methods for constructing protein structure models from 3D electron microscopy maps

Abstract

Protein structure determination by cryo-electron microscopy (EM) has made significant progress in the past decades. Resolutions of EM maps have been improving as evidenced by recently reported structures that are solved at high resolutions close to 3Å. Computational methods play a key role in interpreting EM data. Among many computational procedures applied to an EM map to obtain protein structure information, in this article we focus on reviewing computational methods that model protein three-dimensional (3D) structures from a 3D EM density map that is constructed from two-dimensional (2D) maps. The computational methods we discuss range from de novo methods, which identify structural elements in an EM map, to structure fitting methods, where known high resolution structures are fit into a low-resolution EM map. A list of available computational tools is also provided.

Keywords: 3D Zernike descriptor; 3DZD; CATH; Class, Architecture, Topology, Homologous superfamily. Acronym for the CATH protein structure database; Computational algorithm; DEN; Deformable Elastic Network; EM; EMDB; ENM; Elastic Network Model; Electron Microscopy Data Bank; Electron density map; Electron microscopy; MC; MD; MDFF; Macromolecular structure modeling; Monte Carlo; NMA; NMFF; NMR; Normal Mode Flexible Fitting; Nuclear Magnetic Resonance; PDB; Protein Data Bank; RMSD; Root Mean Square Deviation; SVM; Structure fitting; electron microscopy; molecular dynamics; molecular dynamics flexible fitting; normal mode analysis; support vector machine.

Figure 1

The growth in the number of EM Maps in the Electron Microscopy Data Bank. The total number of entries at each year (horizontal axis), starting in 2002, is shown as a continuous line with filled circles as markers (left vertical axis). Additionally, a continuous line with unfilled circles shows the number of entries deposited each year (right vertical axis).

Figure 2

Distribution of EM map resolutions in the EMDB. Entries are eliminated if the resolution is not provided.

Figure 3

Steps involved in constructing structure models of proteins from an EM map. The first steps involve experiments such as sample preparation and single-particle cryo-EM data collection. Once a 3D EM map is constructed, computational methods are applied to build structural models, which range from determining secondary structure elements to rigid-fitting and flexible fitting approaches. (A) α-helices start to be identifiable in an EM map if its resolution is 10 Å or higher and they can be clearly identified at 6 Å resolution, while β-sheets can be identified in a map at around 5 Å. To follow this route in the diagram the input EM maps should meet the resolution. (B) In order to perform structural fitting to a map the user needs to have atomic-detailed models of the subunits to fit. These can come either from X-ray, NMR spectroscopy, or computational modeling. (C) Some methods use the identified SSEs as input to their rigid fitting algorithm. (D) Coarse-grain models that provide only a backbone trace can be directly derived from identified SSEs in the EM map.

Figure 4

Graphical summary of different scoring terms used in the EM fitting process.