Proteomdaki protein tümleşiklerini modellemek

Most (if not all) proteins function when associated in multimolecular assemblies. An important aim of structural biology is to attain the structures of protein assemblies at the atomic scale. Experimentally, structures are increasingly available but many are still missing or incomplete. Some experimental methods provide high resolution data for small proteins and some provide low resolution data for large proteins. Computational approaches can help bridge this resolution gap. They are needed to determine structural data of multi-molecular protein assemblies at atomic scale. Existing computational methods have made substantial progress toward this aim; however, current approaches are still limited. Some involve manual adjustment of experimental data; some are automated docking methods, which are computationally expensive and not applicable to large-scale proteome studies; still others exploit the symmetry of the complexes, thus they are not applicable to non-symmetrical complexes. Our study aims to take steps toward overcoming these limitations. We have developed a strategy to construct protein assemblies computationally based on binary interactions predicted by a motif-based protein interaction prediction tool, PRISM (PRotein Interactions by Structural Matching). PRISM predicts pair-wise interactions; here we take a step toward multimolecular assemblies, which reflects the more prevalent cellular scenarios. This method is able to construct homo-/hetero-complexes and symmetric/asymmetric complexes without a limitation on the number of components, considers conformational changes and is applicable to large-scale studies. We modeled a benchmark set of various protein assemblies starting from the unbound forms and obtained successful predictions (0.5 - 5.6 Å). Moreover, we exploit different conformations of the proteins available in the Protein Data Bank (PDB) to consider protein flexibility in modeling protein assemblies. We could increase the success in prediction of binary interactions from 27% to 67% and obtained higher accuracy in modeling protein assemblies by exploiting alternative conformations. Furthermore, we modified our method to exploit electron microscopy (EM) density maps to eliminate improper structures. Filtering structures through EM data prevents assembly construction based on wrong structures and saves computational time. We successfully modeled protein assemblies using EM data, most of them with RMSD less than 5 Å and correlation in density maps are close to or higher than 0.8. Comparing our results with other methods' showed higher accuracy in our interface predictions. We present the methods, illustrate their results, and highlight the current limitations.