Integrated, rational molecular replacement

摘要

In the past 50 years, molecular replacement (MR) has grown from being a way to shortcut the occasional determination of identical or closely related structures to being the route to the majority of protein crystal structures. The 2019 CCP4 Study Weekend on Molecular Replacement, held on 9-10 January 2019 in Nottingham, UK, looked briefly at its long history but focused on recent developments that have helped to push its reach to a much greater range of structures. The special issue at https://journals.iucr.org/ special_issues/2020/CCP42019/ contains contributions from the majority of the speakers at the meeting. After an overview of the history of molecular replacement by Eleanor Dodson, Airlie McCoy talked about the development of likelihood-based MR, and how the use of graph databases will allow even more sophisticated branched searches testing different structural hypotheses. The likelihood approach benefits from accounting for the effects of errors, which must be estimated in advance for structural models. Bjorn Wallner discussed methods to accurately estimate the local errors in computational models, which can be used to weight the relative contributions of different parts of these models. Kaushik Hatti focused on global errors, discussing how these can be estimated from a combination of factors including sequence similarity scores, model size, and model quality scores. Randy Read examined the connection between likelihood scores and measures of information gain, showing that information gain can be used to choose the data that will contribute usefully to MR calculations.