Computational Chemical and Structural Biology
RosettaEPR: Developing Novel Protein Structure Prediction Methods Using Sparse Experimental EPR Data
Many key proteins, especially membrane proteins, continue to evade structure determination by traditionally used techniques, such as x-ray crystallography and NMR because they are often either too large for effective NMR experiments or are difficult to crystallize. Site-directed spin-labeling (SDSL), coupled with electron paramagnetic resonance (EPR), is an alternative approach that has few limitations on the size of the protein, which can be studied in its native environment. SDSL-EPR can also be used to determine residue exposure to membrane or solvent and intra-molecular distances of up to 80Å. Despite these advantages, SDSL-EPR cannot directly yield high-resolution structures. Additionally, the structural information obtained from EPR is limited due to the sparseness of the data, and the flexibility of the spin-label linking arm complicates the conversion of experimental data to structural restraints. We hypothesize that, in the absence of a high-resolution structure, SDSL-EPR experimental data with computational modeling can circumvent these problems. We therefore aim to develop a novel approach to membrane protein structure determination by combining SDSL-EPR experimental data with the protein structure prediction program Rosetta. It has recently been demonstrated that the Rosetta protein folding algorithm, in conjunction with sparse EPR experimental restraints, is capable of generating atomic-detail models of T4-lysozyme and aA-crystallin with RMSDs of 1.0Å and 2.6Å, respectively (Alexander et al, Structure, 2008). RosettaEPR will build on these results by incorporating EPR knowledge-based potentials to be used directly during protein folding. Further, Membrane proteinss will be modeled by employing both EPR restraints and the membrane protein-specific scoring functions in RosettaMembrane.