Jean Garnier
Unite d'ingenierie des Proteines, Biotechnologies INRA, 78352 Jouy-en-Josas, Cedex, France

1- Program SIMPA(J. Levin and J. Garnier, Biochim. Biophys. Acta, 1988, 28, 1177-92) when an homologous structure is known. SIMPA (SIMilarity Peptide Analysis) program is based on sequence similarity between a stretch of amino acids (17 amino acid long) of the test sequence and the sequences in a data base of proteins of known structure. Cross validated Q3 is 86% when the data base contains an homologous protein of known structure, if not the accuracy drops to 63-65%.

2- Predictions program SIMPA (J. Levin & J. Garnier, Biochim. Biophys. Acta, 1988, 28, 1177-92) with multiple alignments of homologous sequences (J. Levin et al. Protein Eng., 1993, 6, 849-54). When no homologous structure is known but homologous sequences are known, the cross validated Q3 can be improved from 63-65% to 69-70% by averaging the predictions at each position of a multiple alignment of the known homologous sequences (program CONSENSUS).

3- Program COMBINE (V. Biou et al. Prot. Eng. 1988, 2, 185-191) with multiple alignments of homologous sequences. COMBINE is an expert program of three secondary structure prediction algorithms: GOR III, SIMPA and Bit Pattern. This latter prediction program is based on hydrophobicity patterns and has a cross validated Q3 of 59%. COMBINE cross-validated Q3 is about 65%. COMBINE can be associated to CONSENSUS to yield an accuracy of 69-70% (V. Di Francesco, P.J. Munson and Garnier J., 28th Hawaii International Conference on System Science, 1995.

Asilomar Conference home page