Proteins are the 'workhorse' molecules of life, they participate in nearly every activity that cells carry out. It follows that understanding protein structure and function is essential to understanding life processes, and how to control or modify them. Biochemistry and biophysics experiments give the most accurate data on protein structure and function, but the experiments are often expensive and too specialized for many of the cell and molecular biologists focused on a particular interesting protein. This means that reliable computational predictions of protein structure and function are in high demand. These techniques are also specialized but can be automated, which is the focus of this project, which aims to develop an integrated platform for high-resolution protein structure prediction and structure-based function annotation that is accessible from the Web. This resource will significantly enhance studies of individual proteins as well as processes in cellular biology and other biological sciences. Through the collaboration of the two institutions, students at NCAT will learn state of the art high performance computing methods, and workshops at both institutions will provide greater understanding of the capabilities of the new resource.
Proteins are complex components of biological systems, and studies on their structure and function often require multiple approaches to measurement or modeling. Many of the advanced computer algorithms used in this modeling are highly specialized, involving a number of complicated processes for each aspect of the protein modeling. Biologists whose primary interest is the final result often cannot determine which algorithm or pipeline to choose, how to enter parameters, or how to interpret the resulting models. While continuing to improve the accuracy of the core algorithms in protein structure prediction and structure-based function annotation, this project will also make improvements to domain parsing and assembly, to improve the quality of complex protein structure and function modeling. Another major focus of this project is to develop new protocols that automatically guide protein targets to the most suitable pipelines. In conjunction with this there will be new confidence scoring systems, both global and local, to assist biological users as they interpret the modeling results. In addition, advanced parallel computing and graphic processor unit techniques will be implemented in order to accelerate the pipelines and reduce user's waiting time. New opportunities will be made for improving educational outcomes, in particular for women and minority students, in both University of Michigan and the North Carolina A&T State University. The on-line protein modeling system will be accessible to the community at http://zhanglab.ccmb.med.umich.edu.
