These tools will no longer be maintained as of December 31, 2024. Archived website can be found here. PubMed4Hh GitHub repository can be found here. Contact NLM Customer Service if you have questions.


Pubmed for Handhelds

PUBMED FOR HANDHELDS

Search MEDLINE/PubMed

  • Title: Reconsidering complete search algorithms for protein backbone NMR assignment.
    Author: Vitek O, Bailey-Kellogg C, Craig B, Kuliniewicz P, Vitek J.
    Journal: Bioinformatics; 2005 Sep 01; 21 Suppl 2():ii230-6. PubMed ID: 16204110.
    Abstract:
    MOTIVATION: Nuclear magnetic resonance (NMR) spectroscopy is widely used to determine and analyze protein structures. An essential step in NMR studies is determining the backbone resonance assignment, which maps individual atoms to experimentally measured resonance frequencies. Performing assignment is challenging owing to the noise and ambiguity in NMR spectra. Although automated procedures have been investigated, by-and-large they are still struggling to gain acceptance because of inherent limits in scalability and/or unacceptable levels of assignment error. To have confidence in the results, an algorithm should be complete, i.e. able to identify all solutions consistent with the data, including all arbitrary configurations of extra and missing peaks. The ensuing combinatorial explosion in the space of possible assignments has led to the perception that complete search is hopelessly inefficient and cannot scale to realistic datasets. RESULTS: This paper presents a complete branch-contract-and-bound search algorithm for backbone resonance assignment. The algorithm controls the search space by hierarchically agglomerating partial assignments and employing statistically sound pruning criteria. It considers all solutions consistent with the data, and uniformly treats all combinations of extra and missing data. We demonstrate our approach on experimental data from five proteins ranging in size from 70 to 154 residues. The algorithm assigns >95% of the positions with >98% accuracy. We also present results on simulated data from 259 proteins from the RefDB database, ranging in size from 25 to 257 residues. The median computation time for these cases is 1 min, and the assignment accuracy is >99%. These results demonstrate that complete search not only has the advantage of guaranteeing fair treatment of all feasible solutions, but is efficient enough to be employed effectively inpractice. AVAILABILITY: The MBA(2) software package is made available under an open-source software license. The datasets featured in the Results section can also be obtained from the contact author.
    [Abstract] [Full Text] [Related] [New Search]