Geometric sieving: Automated distributed optimization of 3D motifs for protein function prediction

Brian Y. Chen, Viacheslav Y. Fofanov, Drew H. Bryant, Bradley D. Dodson, David M. Kristensen, Andreas M. Lisewski, Marek Kimmel, Olivier Lichtarge, Lydia E. Kavraki

Research output: Chapter in Book/Report/Conference proceedingConference contribution

11 Scopus citations

Abstract

Determining the function of all proteins is a recurring theme in modern biology and medicine, but the sheer number of proteins makes experimental approaches impractical. For this reason, current efforts have considered in silico function prediction in order to guide and accelerate the function determination process. One approach to predicting protein function is to search functionally uncharacterized protein structures (targets), for substructures with geometric and chemical similarity (matches), to known active sites (motifs). Finding a match can imply that the target has an active site similar to the motif, suggesting functional homology. An effective function predictor requires effective motifs - motifs whose geometric and chemical characteristics are detected by comparison algorithms within functionally homologous targets (sensitive motifs), which also are not detected within functionally unrelated targets (specific motifs). Designing effective motifs is a difficult open problem. Current approaches select and combine structural, physical, and evolutionary properties to design motifs that mirror functional characteristics of active sites. We present a new approach, Geometric Sieving (GS), which refines candidate motifs into optimized motifs with maximal geometric and chemical dissimilarity from all known protein structures. The paper discusses both the usefulness and the efficiency of GS. We show that candidate motifs from six well-studied proteins, including α-Chymotrypsin, Dihydrofolate Reductase, and Lysozyme, can be optimized with GS to motifs that are among the most sensitive and specific motifs possible for the candidate motifs. For the same proteins, we also report results that relate evolutionarily important motifs with motifs that exhibit maximal geometric and chemical dissimilarity from all known protein structures. Our current observations show that GS is a powerful tool that can complement existing work on motif design and protein function prediction.

Original languageEnglish (US)
Title of host publicationResearch in Computational Molecular Biology - 10th Annual International Conference, RECOMB 2006, Proceedings
PublisherSpringer-Verlag
Pages500-515
Number of pages16
ISBN (Print)3540332952, 9783540332954
DOIs
StatePublished - 2006
Externally publishedYes
Event10th Annual International Conference on Research in Computational Molecular Biology, RECOMB 2006 - Venice, Italy
Duration: Apr 2 2006Apr 5 2006

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3909 LNBI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference10th Annual International Conference on Research in Computational Molecular Biology, RECOMB 2006
Country/TerritoryItaly
CityVenice
Period4/2/064/5/06

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Geometric sieving: Automated distributed optimization of 3D motifs for protein function prediction'. Together they form a unique fingerprint.

Cite this