Premium
Identification of Short Turn Motifs in Proteins Using Sequence and Structure Fingerprints
Author(s) -
Wintjens René T.,
Rooman Marianne J.,
Wodak Shoshana J.
Publication year - 1994
Publication title -
israel journal of chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.908
H-Index - 54
eISSN - 1869-5868
pISSN - 0021-2148
DOI - 10.1002/ijch.199400030
Subject(s) - dihedral angle , motif (music) , chemistry , computational biology , structural motif , sequence motif , turn (biochemistry) , sequence (biology) , protein structure , protein family , pattern recognition (psychology) , artificial intelligence , biochemistry , computer science , biology , gene , hydrogen bond , physics , molecule , organic chemistry , acoustics
Families of 20‐residue turn motifs are identified in a dataset of known protein structures using a fully automatic classification procedure that relies on dihedral angle values and atomic root mean square deviations of the polypeptide backbone. Four of the identified motifs, a novel αα connection, and three well‐known αα, βα, and ββ motifs, are used to investigate the possibility of identifying sequences that adopt these motifs in a library of 20‐residue sequence segments from the protein dataset. To this end, several types of fingerprints are derived for individual members of each turn motif family, and for each family as a whole. These fingerprints represent the sequence conservation among family members, or different reduced descriptions of the protein 3D structure that consider the backbone conformation or the tertiary interactions in the context of the parent proteins. All sequence segments in the library are successively mounted onto the fingerprints, without allowing gaps, and the energy of each mount is evaluated using effective potentials derived from known protein structures. The results show that the ability to recognize native sequences associated with a turn motif is improved when different types of fingerprints are combined, and that it fluctuates significantly according to the specific turn family considered. Overall, however, this ability remains rather limited to a level which is much below the native recognition performance generally achieved for full proteins. The fingerprints and their associated potentials are nevertheless found to be quite effective in generating subsets of solutions that are significantly enriched for the correct sequence‐structure combinations. This may have very useful applications in protein folding simulations and in homology modeling.