DAQplugin: AI-based Interactive Residue-Wise Validation of Protein Models for Cryo-EM maps
Microscopy and Microanalysis · 27.07.2026 · 32 (Supplement_1)
DOI: 10.1093/mam/ozag053.450 ↗Kurz gefasst
Fachartikel in der Zeitschrift Microscopy and Microanalysis, online erschienen am 27.07.2026. Verfasst von Genki Terashi und 2 weiteren Autor:innen. Die Arbeit nennt 9 Literaturangaben.
Abstract (Originalsprache)
The rapid growth of cryogenic electron microscopy (cryo-EM) has enabled structure determination of increasingly complex macromolecular assemblies at near-atomic resolution. However, even in high-resolution maps, local variations in density quality often lead to amino-acid misassignment, including sequence register shifts and incorrectly traced backbones. Such errors are difficult to detect using conventional validation metrics, which primarily evaluate global or local map-model agreement (such as cross correlation [1] and Q-score [2]), or stereochemical correctness [3, 4]. To address this challenge, we previously developed the deep-learning-based amino-acid-wise model quality (DAQ) score [5], which evaluates the likelihood between local cryo-EM density features and residue assignments in protein structure models. DAQ has been used as a residue-level validation tool for identifying potential misassignments in deposited and newly built cryo-EM models [6, 7]. During structure modeling and fitting, we further recognized that an interactive validation score would enable users to assess model quality while actively refining structures. Currently, however, there are no tools that provide real-time, deep-learning based evaluation of protein model quality against cryo-EM maps during the modeling process. Here, we present DAQplugin, an interactive implementation of the DAQ score integrated into ChimeraX [8]. DAQplugin provides real-time visualization of residue-level model quality directly on atomic structures. It also provides a simple one-click installation package. By using this plugin, users can immediately identify and correct local modeling errors in ChimeraX’s graphical interface. Figure 1 illustrates DAQ-guided scoring and model refinement for SARS-CoV-2 B.1.1.7 Spike Glycoprotein trimer (EMD-31563, resolution 3.70 Å; PDB 7FET). In the deposited model, DAQ scores indicate that residues 592-616 of chain B have negative values, suggesting a high possibility of incorrect amino-acid assignments (Fig. 1a). To refine this region, we first applied a two-residue shift in residues 593-603 toward the C-terminus, guided by DAQ scores and interactive molecular-dynamics–based refinement using ISOLDE [9]. This model refinement resulted in a clear improvement of DAQ scores (Fig. 1b). Notably, DAQ scores were recomputed and visualized in real time during the refinement process. Next, residues 604-616 were shifted by two residues toward the N-terminus, leading to a corrected placement of a β-hairpin motif (Fig. 1c). In the final refined model, the β-hairpin region exhibits uniformly positive DAQ scores, indicating improved local agreement with the cryo-EM density (Fig. 1d). This example highlights how DAQplugin can help to intuitively detect and correct sequence register errors during model refinement. In this presentation, we will demonstrate how DAQplugin can be used in real time during model building and refinement for cryo-EM maps. We will also discuss DAQplugin’s utility as an interactive validation tool for improving the accuracy of protein structure models. DAQplugin is publicly available at https://cxtoolshed.rbvi.ucsf.edu/apps/chimeraxdaqplugin . Examples of DAQ-guided model refinement using DAQplugin and ISOLDE. Examples of DAQ scoring and model refinement for EMD-31563 (PDB 7FET). (a) DAQ scoring results visualized in ChimeraX. Negative and positive DAQ scores are shown in red and blue, respectively. Ligand molecules (NAG) are shown in green. The box highlights regions with negative DAQ scores. (b) Refinement of residues 593-603 (highlighted in green) by applying a two-residue shift. (c) Refinement of residues 604–616 (highlighted in green) by an additional two-residue shift. (d) Final refined model after DAQ-guided corrections and interactive refinement in ISOLDE.
Zitieren
APA
Terashi, G., Zhu, H., & Kihara, D. (2026). DAQplugin: AI-based Interactive Residue-Wise Validation of Protein Models for Cryo-EM maps. Microscopy and Microanalysis, 32(Supplement_1). https://doi.org/10.1093/mam/ozag053.450
Vancouver
Terashi G, Zhu H, Kihara D. DAQplugin: AI-based Interactive Residue-Wise Validation of Protein Models for Cryo-EM maps. Microscopy and Microanalysis. 2026;32(Supplement_1). doi:10.1093/mam/ozag053.450
BibTeX
@article{Terashi2026,
author = {Terashi, Genki and Zhu, Han and Kihara, Daisuke},
title = {DAQplugin: AI-based Interactive Residue-Wise Validation of Protein Models for Cryo-EM maps},
journal = {Microscopy and Microanalysis},
year = {2026},
volume = {32},
number = {Supplement_1},
doi = {10.1093/mam/ozag053.450}
}
EUVIMED