Systematic construction and analysis of co-expression networks for identification of functional modules and cis-regulatory elements

Jianhua Ruan, Joseph Perez, Brian Hernandez, Garry Sunter, Valerie M. Sponsel

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

1 Scopus citations

Abstract

Gene co-expression networks have been used successfully for discovering biological relationships among genes on a wholegenome scale, such as predicting gene functional modules and cis-regulatory elements. However, those networks are often constructed in an ad hoc manner, and various methods for network construction and analysis have not been fully evaluated and compared. In this study, we propose a method for constructing gene co-expression networks based on mutual k-nearest-neighbor graphs (mKNN), and compare it with two widely used approaches: threshold-based approach and asymmetric k-nearest-neighbor graph approach (aKNN). We show that mKNN is more robust with respect to the presence of experimental noise and scatter genes, and is less sensitive to parameter variations. Furthermore, we propose a topology-based criterion to guide the selection of the optimal parameter for mKNN, and combine the method with a modularity-based community discovery algorithm to predict functional modules. We evaluate the method on both synthetic and real microarray data. On synthetic data, our method, which does not require any user-tuned parameters, is superior to several popular methods in recovering the embedded modules. Using the yeast stress-response microarray data, we show that the overall functional coherence of the modules predicted by our method using the automatically determined parameters is close to optimal. Finally, we apply the method to study a large collection of gene expression microarray data in Arabidopsis thaliana. Remarkably, with our simple method, we have found many functional modules that are much more significant than those reported by previous studies on the same data set. In addition, we are able to predict cis-regulatory elements for the majority of the functional modules, and the association between the cis-regulatory elements and the functional modules can often be confirmed by existing knowledge.

Original languageEnglish (US)
Title of host publicationProceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
PublisherAssociation for Computing Machinery
Pages15-24
Number of pages10
ISBN (Electronic)9781605583020
StatePublished - Jan 1 2010
Externally publishedYes
Event9th International Workshop on Data Mining in Bioinformatics, BIOKDD 2010, Held in Conjunction with 16th ACM SIGKDD Conference on Knowledge Discovery and Data Mining - Washington, United States
Duration: Jul 25 2010Jul 28 2010

Publication series

NameProceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

Conference

Conference9th International Workshop on Data Mining in Bioinformatics, BIOKDD 2010, Held in Conjunction with 16th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
CountryUnited States
CityWashington
Period7/25/107/28/10

Keywords

  • Bioinformatics
  • Cis-regulatory element
  • Co-expression network
  • Functional module
  • Microarray

ASJC Scopus subject areas

  • Software
  • Information Systems

Fingerprint Dive into the research topics of 'Systematic construction and analysis of co-expression networks for identification of functional modules and cis-regulatory elements'. Together they form a unique fingerprint.

Cite this