File(s) under permanent embargo
Iteratively predict protein functions from protein-protein interactions
Current similarity-based approaches of predicting protein functions from protein-protein interaction (PPI) data usually make use of available information in the PPI network to predict functions of un-annotated proteins, and the prediction is a one-off procedure. However the interactions between proteins are more likely to be mutual rather than static and mono-directed. In other words, the un-annotated proteins, once their functions are predicted, will in turn affect the similarities between proteins. In this paper, we propose an innovative iteration algorithm that incorporates this dynamic feature of protein interaction into the protein function prediction, aiming to achieve higher prediction accuracies and get more reasonable results. With our algorithm, instead of one-off function predictions, functions are assigned to an unannotated protein iteratively until the functional similarities between proteins achieve a stable state. The experimental results show that our iterative method can provide better prediction results than one-off prediction methods with higher prediction accuracies, and is stable for large protein datasets.
History
Title of book
Information technology and agricultural engineeringSeries
Advances in intelligent and soft computing; v.134Chapter number
91Pagination
771 - 778Publisher
Springer-Verlag Berlin HeidelbergPlace of publication
Berlin, GermanyPublisher DOI
ISSN
1867-5662eISSN
1867-5670ISBN-13
9783642275364Language
engNotes
first published in 2010 in CMBB 2010 : Proceedings of the First International Conference on Cellular, Molecular Biology, Biophysics and Bioengineering, Qiquihar, China , 25-26 Dec. 2010.Publication classification
B1.1 Book chapterCopyright notice
2012, SpringerExtent
117Editor/Contributor(s)
E Zhu, S SambathUsage metrics
Categories
No categories selectedLicence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC