Equipe Systemes moleculaires et biologie structurale, Laboratoire de Mineralogie Cristallographie, Universite Paris 6, CNRS URA 09, Case 115, 75252 Paris cedex 05, France.
Jacques.Chomilier@lmcp.jussieu.fr
From a data base of 1500 loops from 3 to 8 residues, it has been proposed a classification allowing both to discriminate between length, as well as to separate structural families of common length. This relies on the use of two parameters ; the distance from the first to the last Ca of the loop (L1Ln), and the distance from this axis to the center of mass of the loop backbone (dG). In all sufficiently populated families, conformation and sequence signatures can be deduced every time the conformation and sequence frequencies at some position are higher than if they were equaly distributed over all families at that position. This is thus an indication of a relation between sequence and structure in loops. These findings can be used for molecular modeling. A profile is derived from the occurrence of the residues in the various families, resulting in a score allowing to attribute a structural family to a loop of unknown structure. This is performed with the only information of the sequence of the loop to be modelled. On a short test set, the percentage of correct assignment is in the range 60-70%. These results have been partially published in J. Mol. Biol. (1996) 259 855-872.