Dr. Frank Alber
Department of Biopharmaceutical Sciences
University of California, San Francisco
San Francisco, CA

Monday, September 13, 2004
3:00 pm (CT)
3269 Beckman Institute

Abstract

Protein assemblies are studied using a variety of experimental techniques that provide data of various scales, resolutions and error modalities. Potential data sources include X- ray crystallography of the subunits, low-resolution electron microscopy of the assembly, subunit proximities by affinity chromatography and information on direct interactions of subunits from chemical cross-linking etc. In many cases a single specific method can provide only partial information on some structural aspects of the assembly and in general, none of these partial data sets is sufficient on its own to reconstruct the complete 3D model of the assembly. In order to gain a comprehensive picture of the assembly structure we need to integrate the disparate data of many such sources into a common framework. We express this data integration as an optimization problem in which we i) define a representation of the protein assembly, ii) express all input information as spatial restraints in a scoring function and iii) perform an optimization of the scoring function and sample stochastically many configurations of the protein assembly that are consistent with all available input data. This approach ensures that we maximize the utility of the available data and get the best possible model(s) at its appropriate resolution. This approach will be illustrated by the low-resolution modeling of the yeast nuclear pore complex.

Tea and coffee will be served in R3151 Beckman Institute at 2:15pm.