Unité de Bioinformatique génomique et structurale



The PoPMuSiC program for computer-aided design of single-site mutations evaluates the changes in stability of a given protein or peptide under all possibles mutations

See PoPMuSiC's WebPage


PRELUDE & FUGUE predict the local structure of a protein in terms of backbone dihedral angle domains, identify sequence regions that form early during folding, and locate structural weaknesses, defined as regions whose sequence is not optimal with respect to the tertiary fold.

See PRELUDE & FUGUE's Web Page


SODa predicts whether or not a sequence is an Fe/Mn superoxide dismutase (sod). If so, it predicts in addition its metal ion specificity (Fe, Mn, or cambialistic), and its oligomeric mode (dimeric or tetrameric). SODa also proposes a list of residue mutations likely to modulate or improve the sod-characteristics or the preference for the metal cofactor or oligomerization mode.

See SODa's Web Page


MetaFoRe is a fold recognition program (Kocher et al. , 1994). It proceeds by threading without insertions and deletions the target sequence onto a dataset of 141 known protein structures and by estimating the sequence-structure compatibility on the basis of several database-derived potentials. These potentials used are:

torsion potentials computed from the propensities of amino acids and amino acid pairs to be associated with certain backbone torsion angle domains or pairs of torsion angle domains; they describe local interactions along the chain;

Cm -Cm distance potentials, where the distance between amino acids is computed between average side chain centroids called Cm; they essentially describe non-local interactions along the chain and are dominated by hydrophobic forces;

hydrophobictity potentials, computed from the propensities of amino acids to expose a certain fraction of their surface to the solvent.

SoFi and SoFiSt

The automatic structure alignment programs SoFi and SoFiSt (Boutonnet et al., 1995) determine the optimal alignments between two protein tertiary structures, using as the sole similarity measure the root mean square deviation (rms) of N, Ca, C and O backbone atoms after coordinate superposition. The algorithm comprises three main steps:

(1) identifying polypeptide segments with similar backbone conformations in both 3D-structure

Given two proteins that one wishes to align, a fixed segment length is specified for SoFi while only secondary structures are taken into account for SoFiSt. Then all overlapping segments (short segments or secondary structures) of one protein are successively superimposed onto those of the other protein, each superposition matching a single segment from each protein and the resulting rms is computed using the standard algorithm of Kabsch. Among all superimposed segments pairs, only whose which superimpose with an rms below a certain threshold defined by the user are retained.

(2) the use of a multiple linkage clustering (MLC) algorithm to find combinations of these segments which yield optimal global alignments

To obtain global alignments, superimposed segments are combined using a constrained multiple linkage tree-like clustering algorithm. This novel algorithm is a generalization of the hierarchical single linkage classification and its originality resides in allowing each class to be combined N times. So, each of the pairs of segments identified in the previous step forms a class on its own. Then all pairs of classes are combined and the backbone rms of the segments they contain are computed. The pair of classes with the lowest rms is joined forming a new class. This procedure is repeated until all classes are joined. To reduce computer time, restrictions are imposed. This constrained MLC generates intertwined trees, from which solutions corresponding to the best alignments are selected, first by eliminating those which align less than a specified fraction of the residues, the latter parameter being imposed by the user. Several structure alignments can usually be obtained for a given pais of proteins, which can be exploited to define automatically the common core or the consensus motif from very diverse members of a protein family.

(3) extending the alignments to include residues outside the initial segments.

The structure alignments obtained by the procedure described above are extended to include residues outside the initial segments or secondary structures. This is done by extending in turn each aligned chain segments, one residue at a time, either at its N- or C-terminus. The corresponding r.m.s.d are computed. The extensions that increase the rms value are rejected and among the other tested extensions, that with the lowest rms value is retained. The process is repeated until every further extension increases the rms

In brief, the algorithm determines the best match between all secondary structures or all short segments of the two fragments yielding the lowest global rms.

null About... (A propos de...) | Crédits | ©2005-2011 J.M. K.