The principal goals of this project are the development of algorithms that allow one to make better con- nections between molecular simulations using force ?elds and the interpretation of crystallographic data on biomolecules. This will involve the following components: Improved models for X-ray scattering from bulk solvent. Most crystallographic re?nements protocols try to place a few ?ordered? solvent (water) molecules into the observed density, and treat the remaining solvent space as arising from a ?at density. Neither part of this model is physcially correct: very few solvent molecules are really ordered, and the bulk is far from ?at. This project will explore alternate methods for describing the scattering contribution from solvent, based on integral equation models and explicit molecular dynamics simulations. This is expected to lead to a more correct description of the solvent environment, which it turn should lead to more accurate electron densities and interpretations for the biomolecules themselves Improved models for biomolecular motion and conformational heterogeneity. The standard model for crys- tallographic re?nement treats motion and heterogenety via optimization of atomic displacement parame- ters (ADPs) and some combination of rigid-body (TLS) motions for certain subsets of the biomolecule, along with the speci?cation of ?alternate conformations? where this can be determined from an examina- tion of the density. The proper description of correlated motions in crystals is likely to be key in obtaining better models for X-ray scattering, but relatively little is known about what motional models are the most physically realistic. Molecular dynamics simulations of super-cells (many unit cells) of biomolecular crys- tals will be used to examine these issues, with particular attention to the question of how well re?ned TLS parameters are likely to correspond to physical motions. Using modern force ?elds to guide X-ray re?nement. My group has collaborated for about two years with the Phenix development team to create a software environment that can use modern force ?elds (as imple- mented, for now, in the Amber simulation package) to complement or replace geometric restraints of the Engh-Huber variety with the forces arising from a periodic molecular mechanics model. The initial version of this will be released in the Spring of 2016, and goes a long way towards automating the process from a user's perspective: running a preparation script and setting useAmber=true is often enough to turn on the new procedure. But much remains to be done: allowing for alternate conformations, ensemble-based re?nement, and incorporation of implicit solvent models in a periodic environment will be top priorities, as will be more extended testing and incorporation of user feedback.
X-ray crystallography is a key tool in structural biology. This project will explore ways to improve the creation and re?nment of atomic models based on di?raction data, including novel methods to model bulk solvent contributions, the testing and development of methods to describe collective motions and disorder, and the integration of these ideas into the Phenix re?nement package. A common theme will be to make better use of insights from molecular dynamics simulations of biomolecular crystals. The improved models should have wide applicability to fundamental biomedical research by providing better understanding of large complexes that may only be described at low resolution and by improving the starting points for structure-based drug discovery and design.
|Cerutti, David S; Debiec, Karl T; Case, David A et al. (2017) Links between the charge model and bonded parameter force constants in biomolecular force fields. J Chem Phys 147:161730|