This article is part of the supplement: Selected papers from the Seventh Asia-Pacific Bioinformatics Conference (APBC 2009)
Research
Computational identification of condition-specific miRNA targets based on gene expression profiles and sequence information
-
* Corresponding author: Zhangjun Fei zf25@cornell.edu
1 Boyce Thompson Institute for Plant Research, Cornell University, Ithaca, NY 14853, USA
2 USDA Robert W. Holley Center for Agriculture and Health, Ithaca, NY 14853, USA
BMC Bioinformatics 2009, 10(Suppl 1):S34 doi:10.1186/1471-2105-10-S1-S34
Published: 30 January 2009Abstract
Background
MicroRNAs (miRNAs) are small and noncoding RNAs that play important roles in various biological processes. They regulate target mRNAs post-transcriptionally through complementary base pairing. Since the changes of miRNAs affect the expression of target genes, the expression levels of target genes in specific biological processes could be different from those of non-target genes. Here we demonstrate that gene expression profiles contain useful information in separating miRNA targets from non-targets.
Results
The gene expression profiles related to various developmental processes and stresses, as well as the sequences of miRNAs and mRNAs in Arabidopsis, were used to determine whether a given gene is a miRNA target. It is based on the model combining the support vector machine (SVM) classifier and the scoring method based on complementary base pairing between miRNAs and mRNAs. The proposed model yielded low false positive rate and retrieved condition-specific candidate targets through a genome-wide screening.
Conclusion
Our approach provides a novel framework into screening target genes by considering the gene regulation of miRNAs. It can be broadly applied to identify condition-specific targets computationally by embedding information of gene expression profiles.