Project 4: Integrating Model Validation and Improvement with the Structure

University Of Calif-Lawrenc Berkeley Lab, Berkeley CA

Investigators

Linked publications & trials

Abstract

Project Summary/Abstract Phenix has already achieved an extremely effective, highly automated, and user-friendly system that addresses the entire structure-solution process for macromolecular crystallography. While work will continue on further improvements to the infrastructure, completeness, and integrated tracking, the primary emphasis going forward will be enabling routine, successful, and accurate solution of difficult structures: low resolution (worse than 3A), cases with weak phase information, and large complexes. At low resolution where fit of the model to the data can only be approximate, it is always necessary to make use of outside information, and the Phenix teams propose to augment that process in several innovative ways. The overall goal of Project IV is to provide dynamic structure validation throughout the Phenix crystallographic software system, to automate improved structure solution and accuracy at all resolutions. Difficult structures will be enabled by developing resolution-tuned libraries and strategies, implementing context-dependent parameters, and diagnosing and avoiding systematic distortions both of electron density maps and of the models built from them. Automated methodologies will be developed for exploiting various types of additional information, primarily explicit all-atom contact analysis, more thorough and guided local conformational sampling directly in refinement, and model-building help from the complementary methodology of computational predictions (collaboratively with members of the Rosetta developer community). In particular, crystallography at low resolution would benefit greatly from reliable predictions capable of working piecewise on subsections of a large structure. Accuracy will be improved across the full range of resolutions by more correct treatment of alternate conformations at high resolution, by integrated tracking and presentation of local validation criteria, automation of many types of corrections, and at low resolution by combining all analyses to achieve an internally consistent, relaxed, and well packed model that stays consistent with the experimental data. These procedures and their automated use in the Phenix protocols should ensure outstanding accuracy for the resulting structures. Relevance: Much of the current excitement within structural biology, and in the related fields such as cell, evolutionary, biomedical, and systems biology that make use of that structural information, is focused on low resolution structures of large molecular machines that perform complex and highly regulated biological functions. The improvements in accuracy that we hope to achieve for these difficult structures will enable the entire biomedical research community to gain rich biological and medical insights from large, low-resolution structures more often.