Three-dimensional protein structure determination is crucial in such fields as structure-based drug design and structural proteomics. An established technique for high-resolution molecular structure determination is nuclear magnetic resonance (NMR). However, full automation of the current NMR structure determination methods are limited by the (1) large number of experiments and long spectrometer time, (2) challenges in assignment of chemical shifts and nuclear Overhauser effect (NOE) peaks, (3) poor performance of current computational methods in sparse data settings. Our lab has proposed novel methods to address the key computational bottlenecks in NMR structure determination. The developed methods minimize the number of required NMR experiments and the amount of human intervention to interpret the data by using global orientation constraints obtained from RDCs to first calculate accurate backbones and then resolve assignments of distance constraints from NOEs and side-chain placements to obtain high-resolution protein structures. These methods were successfully used on experimental NMR data with our collaborators. My current project extends this method by incorporating slightly larger amount of RDC data while reducing the amount of needed NOEs, and extending the provable guarantees on the accuracy of the resulting backbone structures. In this talk I will first overview the general methodology and results obtained in our lab, and then present the project I have been working on under this scope.
Anna Yershova received her Bachelor's degree in Mathematics from Kharkiv National University, Ukraine in 1999, her M.S. in Computer Science from Iowa State University in 2003, and her Ph.D. in Computer Science from University of Illinois in Urbana-Champaign in 2008. Her Ph.D. thesis developed sampling and searching methods in robot motion planning. She is currently a postdoctoral associate at Duke University working in Bruce Donald's Lab on computational structural biology.