Date Published: March 01, 2020
Publisher: International Union of Crystallography
Author(s): Grzegorz Chojnowski, Koushik Choudhury, Philipp Heuser, Egor Sobolev, Joana Pereira, Umut Oezugurel, Victor S. Lamzin.
A novel method for the enhancement of automated protein model building using polypeptide fragments of homologous structures is presented.
Model building is a key step in macromolecular crystallographic structure determination. With the availability of X-ray diffraction data to a resolution of better than 3.0 Å and an initial map of reasonable quality, model building can often be accomplished using automated approaches. The automated tools not only accelerate the model building itself but, more importantly, can also help to avoid subjectivity throughout the density-map interpretation process. The performance of crystallographic model-building methods is reduced at lower resolution owing to the lower information content of the data (Karmali et al., 2009 ▸). For these cases the protein backbone models become fragmented, may contain insertions, deletions or incorrect connections, and may become difficult to assign to the target sequence (Chojnowski et al., 2019 ▸).
With over 130 000 crystal structures currently available in the PDB, it may be possible to find a homologue for many newly crystallized proteins. Indeed, the MR method accounts for almost 80% of the solved structures deposited in the PDB. Apart from assisting in structure solution using MR, homology has been exploited for crystallographic model building and refinement when a sequence assignment is available (van Beusekom, Joosten et al., 2018 ▸; van Beusekom, Touw et al., 2018 ▸; Kovalevskiy et al., 2016 ▸; Nicholls et al., 2012 ▸; Schröder et al., 2010 ▸; Smart et al., 2012 ▸; Headd et al., 2012 ▸). It is intriguing that the majority of model-building tasks that have recently been submitted to the ARP/wARP web service had a homologous structure with a sequence identity of 35% (or greater) already available in the PDB.