LIKELIHOOD BASED PHASING METHODS IN PHASER

University Of Calif-Lawrenc Berkeley Lab, Berkeley CA

Investigators

Linked publications & trials

Abstract

For success in the structural genomics initiatives, it is essential that the methods be as powerful, as reliable,[unreadable] and as automated as possible. In work on our new program, Phaser, we have been applying likelihood-based[unreadable] methods to solving the phase problem of X-ray crystallography by both molecular replacement and[unreadable] experimental phasing. Likelihood-based methods have two advantages in the context of automation. First,[unreadable] they are more powerful and therefore have a greater convergence radius. Second, likelihood is a natural[unreadable] scoring function to measure success in satisfying the data, which makes it an excellent criterion for[unreadable] automated decision-making. We will enhance the automation features that already exist in the molecular[unreadable] replacement and experimental phasing modules of Phaser so that, as much as possible, only the diffraction[unreadable] data and sequence information are needed as input. We will work to increase the sophistication and power[unreadable] of our methods, by taking account of more of the correlations that have been neglected to date in order to[unreadable] reduce the computational demands. We will also work to make our algorithms easily accessible to other[unreadable] Projects within PHENIX, by using the Boost.Python library to make them available from the Python scripting[unreadable] language that underpins PHENIX. This will allow the development of automated pipelines combining[unreadable] algorithms from all the components of PHENIX.[unreadable] With the completion of the human genome, there is a new focus on finding ways to exploit the wealth of[unreadable] information it has revealed. In many cases, knowing the 3D structure of the proteins encoded by the genes[unreadable] is invaluable in understanding how they work together, and the primary method to determine those structures[unreadable] is the method of X-ray crystallography. Our work on applying statistical methods to 3D structure[unreadable] determination is helping to make those methods faster, more automatic and more powerful.