Macromolecular assemblies are the basic functional units of biological cells; they furnish targets for drug design, as deficiencies in their architecture are frequently linked to health problems. The overall goal of the proposed research is the development of computational quantitative fitting tools for electron microscopy (EM) that combine low-resolution image reconstructions of large assemblies with complementary atomic resolution data of individual subunits for routine determination of the large-scale structure of aggregates. Key questions to be addressed include: (i) How can one accurately include experimental geometric constraints in the docking of single molecules? (ii) Are the features present in EM maps of assemblies sufficiently well-defined for a rapid, coarse-grained registration of template structures based on density estimation? (iii) Is a six-dimensional rigid-body search efficient and robust under experimental data limitations such as unaccounted parts in the compared structures? We will use topology representing neural networks for a coarse estimation of density maps and for determining suitable landmarks for the registration of multi-resolution data. Complementary to this indirect approach an exhaustive rigid body search will be performed in reciprocal space using parallel computing architectures (for computational speed) based in part on the gradient of the compared data sets (for accuracy). A computational laboratory supported by this project will be used extensively for software development and for fitting applications in EM. Collaborative efforts will include the refinement of actin filaments and microtubules, and the study of their interactions with motors and with factors promoting their disassembly, and the modeling of elongation factors on the ribosome. The results of these developments will be new computer codes that provide a comprehensible and flexible approach to the multi-resolution modeling of large assemblies. The algorithmic and methodological developments will be distributed freely through the established internet-based mechanisms employed for the Situs docking package.
Showing the most recent 10 out of 45 publications
