PROJECT 2 - MODEL COMPLETION AND VALIDATION

University Of Calif-Lawrenc Berkeley Lab, Berkeley CA

Investigators

Linked publications & trials

Abstract

As part of the PHENIX project we have developed the RESOLVE model-building algorithm, made major[unreadable] advances in density modification procedures, and developed rapid fitting of flexible ligands to electron[unreadable] density maps. We have collaborated with other members of the PHENIX project integrating these algorithms[unreadable] into PHENIX and to developing structure determination Wizards suitable for automated decision-making.[unreadable] We propose to use this foundation of algorithms and software in PHENIX as a basis for automating the[unreadable] completion and validation of macromolecular structures. In collaboration with project I, we will generate[unreadable] highly complete models at moderate resolution. In collaboration with Project IV we will develop templatematching[unreadable] techniques for model-building of nucleic acids. We will create a multi-model representation of the[unreadable] conformations of a macromolecule present in a crystal structure. We will develop our "full-omit" procedure to[unreadable] generate electron density maps that are highly accurate but unbiased by atomic models, and use them in[unreadable] validation procedures. We will develop a two-pass procedure for structure determination, a rapid pass to get[unreadable] a preliminary model and phases, then a second thorough pass making decisions based on correlations of[unreadable] phases to the model phases. We will develop PHENIX Solution objects as a framework for recording all[unreadable] steps carried out during structure determination, allowing repetition comprehensive deposition. We will[unreadable] develop systematic procedures for assessing the quality and characteristics of experimental data, maps and[unreadable] models. We will create a database of quality measures, decisions and their outcomes by carrying out[unreadable] structure determinations using PHENIX structure library and collaborate with Projects I, III and IV to develop[unreadable] machine-learning algorithms for decision-making in structure determination. This work will allow rapid[unreadable] determination and deposition of highly complete models of macromolecules and will be important for[unreadable] understanding biology and improving human health.