PT  - JOURNAL ARTICLE
AU  - Cragnolini, Tristan
AU  - Beton, Joseph
AU  - Topf, Maya
TI  - Cryo-EM structure and B-factor refinement with ensemble representation
AID  - 10.1101/2022.06.08.495259
DP  - 2022 Jan 01
TA  - bioRxiv
PG  - 2022.06.08.495259
4099  - http://biorxiv.org/content/early/2022/06/09/2022.06.08.495259.short
4100  - http://biorxiv.org/content/early/2022/06/09/2022.06.08.495259.full
AB  - Cryo-EM experiments produce images of macromolecular assemblies that are combined to produce three-dimensional density maps. It is common to fit atomic models of the contained molecules to interpret those maps, followed by a density-guided refinement. Here, we propose TEMPy-REFF, a novel method for atomic structure refinement in cryo-EM density maps. By representing the atomic positions as components of a mixture model, their variances as B-factors, and a model ensemble description, we significantly improve the fit to the map compared to what is currently achievable with state-of-the-art methods. We validate our method on a large benchmark of 366 cryo-EM maps from EMDB at 1.8-7.1Å resolution and their corresponding PDB assembly models. We also show that our approach can provide newly-modelled regions in EMDB deposited maps by combining it with AlphaFold-Multimer. Finally, our method provides a natural interpretation of maps into components, allowing us to accurately create composite maps.Competing Interest StatementThe authors have declared no competing interest.