Adaptive Smith-Waterman residue match seeding for protein structural alignment

Abstract

The POLYFIT rigid-body algorithm for automated global pairwise and multiple protein structural alignment is presented. Smith-Waterman local alignment is used to establish a set of seed equivalences that are extended using Needleman-Wunsch dynamic programming techniques. Structural and functional interaction constraints provided by evolution are encoded as one-dimensional residue physical environment strings for alignment of highly structurally overlapped protein pairs. Local structure alignment of more distantly related pairs is carried out using rigid-body conformational matching of 15-residue fragments, with allowance made for less stringent conformational matching of metal-ion and small molecule ligand-contact, disulphide bridge, and cis-peptide correspondences. Protein structural plasticity is accommodated through the stepped adjustment of a single empirical distance parameter value in the calculation of the Smith-Waterman dynamic programming matrix. Structural overlap is used both as a measure of similarity and to assess alignment quality. Pairwise alignment accuracy has been benchmarked against that of 10 widely used aligners on the Sippl and Wiederstein set of difficult pairwise structure alignment problems, and more extensively against that of Matt, SALIGN, and MUSTANG in pairwise and multiple structural alignments of protein domains with low shared sequence identity in the SCOP-ASTRAL 40% compendium. The results demonstrate the advantages of POLYFIT over other aligners in the efficient and robust identification of matching seed residue positions in distantly related protein targets and in the generation of longer structurally overlapped alignment lengths. Superposition-based application areas include comparative modeling and protein and ligand design. POLYFIT is available on the Web server at http://polyfit.insa-toulouse.fr.

An Atomistic Statistically Effective Energy Function for Computational Protein Design

“…Structural analyses of protein domains and protein–ligand complexes were performed using the Panorama program interfaced with the residue libraries . Hydrogen atoms are built automatically by the program, and atom typing was performed on the basis of the updated atom connectivity information.…”

“…Hydrogen atoms are built automatically by the program, and atom typing was performed on the basis of the updated atom connectivity information. Histidine protein residue protonation states were assigned by previously described hydrogen bond network optimization . Aspartic and glutamic acid residue side chains were assumed to be unprotonated.…”

“…The procedure involved the flagging of interactions of potentially incorrectly assigned ligand atom types in protein complexes using a simple distance-dependent polar atom interaction scoring regime. Crosschecking was performed against residue lists of substructures with known tautomer form preferences in water 124 125 Hydrogen atoms are built automatically by the program, and atom typing was performed on the basis of the updated atom connectivity information. Histidine protein residue protonation states were assigned by previously described hydrogen bond network optimization.…”

“…Histidine protein residue protonation states were assigned by previously described hydrogen bond network optimization. 125 Aspartic and glutamic acid residue side chains were assumed to be unprotonated. Disulfide-bonded cysteine residues were identified using a geometry quality scoring scale.…”