Using the Rosetta algorithm and selected inter-residue distances to predict protein structure



The Rosetta algorithm has had much success in protein structure prediction as demonstrated in the recent Critical Assessment of Protein Structure Prediction (CASP) experiments. For many proteins, Rosetta generates several low root mean square deviation (RMSD) decoy structures but finding the best structure among the decoys can be difficult. Experimental data can be used to aid in the discrimination process. Our protein structure prediction method involves three steps: using the Rosetta algorithm to generate decoys, measuring inter-residue distances, and comparing the measured distances with those calculated in each decoy. Decoys with similar three-dimensional structure will also have several similar inter-residue distances. To develop our search protocol, we determined the optimal number of decoys to generate as well as the minimum number of distance constraints needed to distinguish between the low and high RMSD structures. To test our method, we simulate experimental data by measuring α-carbon distances from the experimentally determined structures of our target proteins. We have employed the Rosetta algorithm to generate decoy sets of different sizes for four target proteins. Our predicted structures ranged in Cα RMSD from 2.4 to 4.6 Å compared with the experimental structures. Using only twenty-five distance constraints, reliable predictions were made. © 2008 Wiley Periodicals, Inc. Int J Quantum Chem, 2008