Modeling cytochrome P450 14 alpha demethylase (Candida albicans) from P450cam.

The tertiary structure of cytochrome P450 14 alpha demethylase--Candida albicans (P450 CA) is modeled on the basis of sequence alignment with two closely related proteins and the crystallographic structure of Pseudomonas putida P450cam. The secondary structure prediction system used combines the information from several algorithms and trains the data to offer an optimized prediction of the known P450cam. The trained algorithm was then used to predict the secondary structure of the other P450 sequences. The prediction of the surface coil regions was aided by an alignment between P450 CA and the homologous sequences P450 14 alpha demethylase--Saccharomyces cerevisiae (66 SD) and P450 14 alpha demethylase--Candida tropicalis (72 SD). The prediction and alignment information was combined to establish an alignment between P450 CA and P450cam, and to assign full secondary structure to the target protein. This secondary structure was folded from the template of P450cam and the predicted structure was relaxed by molecular dynamics. Model checking highlighted minor adjustments in the alignment, correctly orienting hydrophobic and hydrophilic side chains. The model offers explanations for several known experimental results and suggests further investigations that may prove fruitful in understanding the structure and mechanisms of the P450 family (Porter, T.D. and Coon, M.J. Minireview cytochrome P450. J. Biol. Chem. 1991, 266, 13469-13472. Waterman, M.R. Cytochrome P450 cellular distribution and structural considerations. Current Opinion in Structural Biology 1992, 2, 384-387. Aoyama, Y., Yoshida, Y., Sonohdo, Y. and Sato, Y. Structural analysis of the interaction between the side-chain of substrates and the active site of lanosterol 14 alpha demethylase (P450 14DM) of yeast. Biochim. Biophys. Acta 1992, 1122, 251-255.).

Secondary structure prediction for modelling by homology.

An improved method of secondary structure prediction has been developed to aid the modelling of proteins by homology. Selected data from four published algorithms are scaled and combined as a weighted mean to produce consensus algorithms. Each consensus algorithm is used to predict the secondary structure of a protein homologous to the target protein and of known structure. By comparison of the predictions to the known structure, accuracy values are calculated and a consensus algorithm chosen as the optimum combination of the composite data for prediction of the homologous protein. This customized algorithm is then used to predict the secondary structure of the unknown protein. In this manner the secondary structure prediction is initially tuned to the required protein family before prediction of the target protein. The method improves statistical secondary structure prediction and can be incorporated into more comprehensive systems such as those involving consensus prediction from multiple sequence alignments. Thirty one proteins from five families were used to compare the new method to that of Garnier, Osguthorpe and Robson (GOR) and sequence alignment. The improvement over GOR is naturally dependent on the similarity of the homologous protein, varying from a mean of 3% to 7% with increasing alignment significance score.