| Prediction of a protein fold is a big challenge in modern science. Strictly from a geometrical standpoint, there are billions of possible configurations for a single protein. However, experiment shows that for a specific sequence, only a few, closely related conformations are observed. Our objective is to correctly predict the three-dimensional structure of a protein, with a minimum set of a priori knowledge.; John Gunn's research group has developed a protein folding model: TRIP. Based on the knowledge of the secondary structure, this model uses a hierarchical scheme for its conformational search. It guides the generation of thousands of conformations through a series of simulated annealing, Monte Carlo and a genetic algorithm. Filters may be added to any hierarchical construction level in order to select the best conformations in the randomly generated pool.; Possible ways of improving the prediction with Ramachandran map will be discussed here. It will be shown that it is possible to locally improve structures. In addition, with the use of locally enhanced Ramachandran maps, this model may use more than one secondary structure pattern in the same calculation. A special focus on β sheets will be included.; A new algorithm that constructs β sheets from independent β strands will be presented. Its effectiveness will be demonstrated with a three dimensional functional fitting of the resulting β sheets. |
