Journal of M olecular Structure (Theo&m), 232 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPO (1991) 321-336 Elsevier Science Publishers B.V., Amsterdam 321 zyxwvut PREDICTION OF PROTEIN SECONDARY STRUCTURES USING A COMBINED METHOD BA SED ON THE RECOGNITION, LIM AND GARNIER-OSGUTHORPE-ROBSON ALGORITHMS S. THORNTON”, E. SAN FABIAN*,‘, S. FRAGA”, J.M.R. PARKERb and R.S. HODGESb “Department of Chemistry, University of Alberta, Edmonton, Alberta T6G 2G2 (Canada) bDepartment of Biochemistry, University of Alberta, Edmonton, Alberta T6G 2H7 (Canada) (Received 9 July 1990) ABSTRACT The prediction of the secondary structure of a protein, discussed in this paper, is based on the algorithm developed by Lim, extensively modified and coupled with the minimum recognition method developed in this laboratory, as well as the Gamier-Osguthorpe-Robson method. The modified Lim algorithm identifies eight types of pattern as possibly having structural character when given conditions are satisfied. The preliminary prediction of the a! helices and /3 chains is carried out by inspection of the residue sequence, identifying first the (Y helices and then the jl chains. The final prediction is then obtained by combining, according to well-defined rules, the above results with those from the minimum recognition and the Gamier-Osguthorpe-Robson methods. The complete algorithm has been developed through parameterization for 80 proteins with known secondary structure. The usefulness of its predictive power, as a fit step towards the computer simulation of a tertiary structure,becomes evident from the following results: 74% (57% ) of the predicted rxhelices (/I chains) are found in correspondence with corresponding experimen- tal fragments, with a sequence position overlap of 78% (66% ) . INTRODUCTION A theoretical simulation of the tertiary structure of a water-soluble globular protein is feasible in the sense that the necessary methodology and means of computation are available. Stable spatial conformations may be obtained un- der a minimum-energy criterion, using an appropriate potential function, through variation of the dihedral angles. However, the probability of arriving at the native conformation of the peptidic chain is negligible for most proteins because of the large number of degrees of freedom. In addition, the computing cost will be considerable. A successful, a priori prediction of the structural fragments, that is, of the ‘Permanent address: Departamento de Quimica Fisica, Universidad de Alicante, Spain.