Premium
Fold recognition from sequence comparisons
Author(s) -
Koretke Kristin K.,
Russell Robert B.,
Lupas Andrei N.
Publication year - 2001
Publication title -
proteins: structure, function, and bioinformatics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.699
H-Index - 191
eISSN - 1097-0134
pISSN - 0887-3585
DOI - 10.1002/prot.10000
Subject(s) - sequence (biology) , computer science , protocol (science) , computational biology , fold (higher order function) , protein secondary structure , artificial intelligence , pattern recognition (psychology) , biology , medicine , genetics , programming language , pathology , biochemistry , alternative medicine
We applied a new protocol based on PSI‐Blast to predict the structures of fold recognition targets during CASP4. The protocol used a back‐validation step to infer biologically significant connections between sequences with PSI‐Blast E‐values up to 10. If connections were found to proteins of known structure, alignments were generated by using HMMer. The protocol was implemented in a fully automated version (SBauto) and in a version that allowed manual intervention (SBfold). We found that the automated version made 17 predictions for target domains, of which 8 identified the correct fold with an average alignment accuracy of 24% for alignable residues and 43% for equivalent secondary structure elements. The manual version improved predictions somewhat, with 10 of 15 predictions identifying the correct fold with alignment accuracies of 33% for alignable residues and 64% for equivalent secondary structure elements. We describe successes and failures of our approach and discuss future developments of fold recognition. Proteins 2001;Suppl 5:68–75. © 2002 Wiley‐Liss, Inc.