Everything, Everywhere, All at Once: multi-state, multi-dataset, multi-model refinement -Nick Pearce

The process of building and refining accurate macromolecular models remains critical for obtaining high-resolution structural information. Traditional refinement approaches often still involve the time-consuming optimization of a single model against a single experimental data set. However, with increasing automation, especially in the study of weak structural features such as ligand binding in fragment screening experiments, we require methodologies that embrace the complexity inherent in biological systems, leading to a series of models containing multiple states derived from a series of data sets.
We present methods for simplifying the generation and refinement of multi-state models for use in multi-dataset experiments. Using automatically generated structural restraints to aid convergence, we show how weak structural features can be simply incorporated into models, reliably refined en masse, and finally intuitively validated using a self-consistency-based
analysis alongside conventional model validation. This method is applied to fragment screening of the SARS-CoV-2 protease.