Protein model refinement for cryo‐EM maps using AlphaFold2 and the DAQ score. Issue 1 (20th December 2022)

• Record Type: Journal Article
• Authors: Terashi, Genki; Wang, Xiao; Kihara, Daisuke
• Is Part Of: Acta crystallographica. Volume 79:Issue 1(2023)
• Page Start: 10
• Page End: 21
• Publication Date: 2022-12-20
• Subjects: cryo‐electron microscopy -- protein model refinement -- protein model validatation -- computational methods -- AlphaFold2 -- DAQ score; X-ray crystallography -- Periodicals; Crystallography -- Periodicals; Molecular biology -- Periodicals; Molecular structure -- Periodicals; Biomolecules -- Structure -- Periodicals; Cytology -- Periodicals; Biomolecules -- Structure; Crystallography; Cytology; Molecular biology; Molecular structure; X-ray crystallography; Periodicals; 548
• DOI: 10.1107/S2059798322011676
• Languages: English
• ISSNs: 2059-7983

Abstract

Abstract : A new protocol, DAQ‐refine, for evaluating a protein model built from a cryo‐EM map and applying local structure refinement is described. Abstract : As more protein structure models have been determined from cryogenic electron microscopy (cryo‐EM) density maps, establishing how to evaluate the model accuracy and how to correct models in cases where they contain errors is becoming crucial to ensure the quality of the structural models deposited in the public database, the PDB. Here, a new protocol is presented for evaluating a protein model built from a cryo‐EM map and applying local structure refinement in the case where the model has potential errors. Firstly, model evaluation is performed using a deep‐learning‐based model–local map assessment score, DAQ, that has recently been developed. The subsequent local refinement is performed by a modified AlphaFold 2 procedure, in which a trimmed template model and a trimmed multiple sequence alignment are provided as input to control which structure regions to refine while leaving other more confident regions of the model intact. A benchmark study showed that this protocol, DAQ‐refine, consistently improves low‐quality regions of the initial models. Among 18 refined models generated for an initial structure, DAQ shows a high correlation with model quality and can identify the best accurate model for most of the tested cases. The improvements obtained by DAQ‐refine were on average larger than other existing methods.