Eliminating Critical Systematic Errors In Structural Biology With Next-Generation Simulation

项目总结/文摘

PROJECT SUMMARY/ABSTRACT Macromolecular Crystallography (MX) is an established and widely used method for obtaining accurate, high- resolution 3D models of biological molecules, yet MX data contain information that has yet to be unlocked. Single-electron changes can be clearly visible at resolutions as low as 3.5 Å if systematic errors can be eliminated. Creating simulation technologies that can account for these errors will have significant impact on three fronts: 1) eliminating the structural changes and other caveats of radiation damage, which ultimately limits the amount of data available from a given sample 2) improving multi-crystal averaging and comparison by capturing and correcting non-isomorphism, which will open the gateway to arbitrary gains in signal-to-noise, 3) discriminating hotly contested alternative interpretations such as the presence or absence of a bound ligand, by creating simulations with more realistic solvent and protein models. To move towards damage-free data from a synchrotron, we will start by implementing a new kind of data collection we call “painting with X-rays” that leverages modern fast-framing detectors to combine the best features of broad-beam and micro-beam technologies: low dose contrast and isolation of the best parts of the crystal. We will then enhance zero-dose extrapolation to handle the rich temporal information made available by finely dividing up the available photons. We will build on our success correcting non-isomorphism in real space into reciprocal space, enabling merging of incomplete data such as XFEL stills into parametric structure factor frameworks. These low-dimensional frameworks will allow selection from a continuum of 3D molecular structures by dialing in desired parameter values, and will also be applied to cases where the parameters are known quantities, such as time-resolved, temperature series, humidity, or other reaction coordinates and variables controlled in an experiment. We will test these framework models against the thousands of non-isomorphous data sets that have been collected at our beamline and report on best practice. To enable robust interpretation of experimental data, we will extend these multi-conformer models with simulation-based solvent models. Our work will create standard protocols for comparing solvent density to alternative interpretations and to quantitatively assess how likely each simulated situation is compared to the real macromolecular crystallography data. In addition to distinguishing between different interpretations of the experimental data, improving solvent models will enhance understanding of how macromolecules influence and interact with other molecules near their surface. Collectively, we expect the benefits of eliminating these critical systematic errors to be transformative to both methods development and functional studies using complimentary structural techniques, such as CryoEM, SAXS, tomography and electron diffraction, especially hybrid methods that combine structural data from multiple sources.