Computation for Structural Biology: Tools to Enable Dynamic 3-D Reconstruction of Time-varying Viral Structures

Purdue University, West Lafayette IN

Investigators

Abstract

Proposal #0098156 Purdue Research Foundation Doerschuk, Peter C A key challenge in computational structural biology is the determination of the 3-D structure of a virus, and especially dynamical changes in 3-D structure which are central to understanding the function of the virus. This information is central to rational design of drugs to combat viral infections and to the use of viruses for other purposes, e.g., as vehicles for the targeted delivery of drugs to specific organs. Solving these problems involves the development, analysis, implementation and use of new algorithms for two numerical computation problems, global optimization and multidimensional quadrature. The investigators compute a 3-D structure by locating the global minimum of a cost which is a function of experimental data and of a predictor, and which quantitates the difference between the data and the prediction of the data. The predictor, whose evaluation requires multidimensional quadrature, is a function of parameters describing the 3-D structure of the virus and any unknown aspects of the data collection process and the minimization is with respect to these parameters. Performance of the approach is limited by the global optimization and quadrature tools and therefore these tools are the foci of this research. Key global optimization issues are exploiting the multi-scale structure of the data and parameters in the cost due to the presence of Fourier transforms and the tradeoff between accuracy and computational expense in the evaluation of the cost due to embedded quadratures. Key multidimensional quadrature issues are the unusual integrands and regions of integration, e.g., to integrate a function of three variables with icosahedral symmetry over the three Euler angles that define a 3-D rotation.