UBC Theses and Dissertations
Prediction program of secondary structure from sequence of proteins according to the method of Chou and Fasman Pham, Anne-Marie
Several methods have been proposed for predicting the secondary structure of proteins. The method of Chou and Fasman (1974a, 1974b, 1978a, 1978b) is relatively simple in theory and reasonably accurate. Unfortunately, the rules of Chou and Fasman are sometimes ambiguous and can be interpreted differently by researchers. Several attempts have been made for computerization of the rules of Chou and Fasman (Argos et al., 1976; Chou and Fasman, 1978b; Dzionara et al., 1977). However, they are for computation of only a portion of the protein secondary structure. The final assignment of the entire structure has to rely on the individual's manipulation. In addition to three separate computer programs for prediction of the α-helix, β-sheet and β-turn structures, a fourth program was written for clarifying overlapping areas between α-helix and β-sheet. Although the predicted structures of 24 proteins with known conformation were in general satisfactory, there were a number of missing areas and boundary values different from X-ray diffraction patterns. In an attempt to improve the accuracy of the prediction, the nucleation rules were modified to emphasize importance of the type and positions of amino acid residues in the region. Furthermore, an extra step for boundary adjustment was added to the search for α-helix and β-sheet regions. This step compared the importance of the boundary conformational parameters and the possible interference of the different conformations at the boundaries of the predicted regions. These modifications produced predicted secondary structures which were in good agreement with the X-ray diffraction patterns and the predicted patterns of Chou and Fasman (1974b, 1978b). The Matthews' coefficient (C) calculated for α-helix and β-sheet were 0.39 or above, meaning that the prediction would be quite useful although there might be one or two helical regions missed or overpredicted. The paired-sample t-test revealed that the values of C[sub α] (P < 0.01) and C[sub β] (P ≤ 0.05) calculated for the present prediction were significantly improved from the values of Chou and Fasman. The computer-assisted technique described in this thesis, therefore, would decrease the discrepancy between the predicted data from different researchers due to the ambiguous interpretations of the rules of Chou and Fasman. The second part of this study involved the application of the program to several food related proteins (bovine serum albumin, α[sub S1]-casein, β-casein, κ-casein, chymosin, α-lactalbumin, α-lactoglobulin, ovalbumin, pepsin and trypsinogen). Although references could not be found for all proteins tested, the results obtained for κ-casein and α-lactalbumin were comparable to those reported by other researchers. Since conformational data have long been recognized as contributing to the information on protein and enzyme functionality, the computerization of the predictive method of Chou and Fasman will definitely be a tool for explaining the protein functionality in food processing.
Item Citations and Data