| Popis: |
The present paper evaluates the effect of several parameters on the accuracy of homology model-built structures of human thrombin. In order to appraise the merits of such a homology model-building study as a practical basis for direct drug-design purposes, this study was limited to the widely used “standard” alignment and loop-modelling methods implemented in the Quanta/CHARMm 3.0 software. The parameters studied were the alignment method, the choice of the template protein, and the molecular dynamics refinement protocol. The resulting models were energy-minimised and compared with the energy-minimised crystal structure of thrombin. It was found that the method of alignment is the dominating parameter: the alignment algorithm giving the most accurate models is one that is based upon the three-dimensional overlap of structurally conserved regions in related proteins. The choice of the template protein structures, which all showed a ± 30% sequence identity with thrombin, turned out to be less important for the accuracy of the final model. In our hands, molecular dynamics simulations at 300 and 1000K failed to improve the predictive value of the models. The overall root-mean-square deviation between the best model and the minimised crystal structure of thrombin amounted to 2.9 A. Detailed comparison of the models with the experimentally determined 3D structure of thrombin leads to the conclusion that although a homology-building study may result in a three-dimensional model that yields a good general idea of the protein structure, rather large structural uncertainties in the proximity of insertions of more than 4 amino acids limit its practical use for direct drug-design efforts. |
