Abstract
microRNAs (miRNAs) are short non-coding RNAs with regulatory functions in various biological processes including cell differentiation, development and oncogenic transformation. They can bind to mRNA transcripts of protein-coding genes and repress their translation or lead to mRNA degradation. Conversely, the transcription of miRNAs is regulated by proteins including transcription factors, co-factors, and messenger molecules in signaling pathways, yielding a bidirectional regulatory network of gene and miRNA expression. We describe here a least angle regression approach for uncovering the functional interplay of gene and miRNA regulation based on paired gene and miRNA expression profiles. First, we show that gene expression profiles can indeed be reconstructed from the expression profiles of miRNAs predicted to be regulating the specific gene. Second, we propose a two-step model where in the first step, sequence information is used to constrain the possible set of regulating miRNAs and in the second step, this constraint is relaxed to find regulating miRNAs that do not rely on perfect seed binding. Finally, a bidirectional network comprised of miRNAs regulating genes and genes regulating miRNAs is built from our previous regulatory predictions. After applying the method to a human cancer cell line data set, an analysis of the underlying network reveals miRNAs known to be associated with cancer when dysregulated are predictors of genes with functions in apoptosis. Among the predicted and newly identified targets that lack a classical miRNA seed binding site of a specific oncomir, miR-19b-1, we found an over-representation of genes with functions in apoptosis, which is in accordance with the previous finding that this miRNA is the key oncogenic factor in the mir-17-92 cluster. In addition, we found genes involved in DNA recombination and repair that underline its importance in maintaining the integrity of the cell.
Highlights
MiRNAs are small endogenous RNAs with a length of about 22 nt with gene regulatory functions and are found in plants and animals [1]
After a filtering step to remove genes and miRNAs that vary little across samples, we used 311 miRNA expression levels and 4,878 gene expression levels to study the functional interactions between both types of molecules
We generated predictions of gene expression in 10-fold cross-validation, and recorded mean squared errors along with all miRNAs identified as predictors
Summary
MiRNAs are small endogenous RNAs with a length of about 22 nt with gene regulatory functions and are found in plants and animals [1]. Unlike other classes of small RNAs, miRNAs undergo a characteristic biogenesis which consists of a transcript folding back on itself to form a distinctive hairpin structure [2]. MiRNAs form a complex with an Argonaute protein, pair with the target mRNA and induce post-transcriptional repression of the gene product [1]. Since more than half of the human protein-coding genes seem to have conserved miRNA pairing sites in their 39-UTR [3], it is difficult to find a biological process or pathway which is not at all influenced by regulation from miRNAs [1]. The second mechanism, which requires less sequence complementarity between mRNA and miRNA, is used more often [1]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
