Removal of AU bias from microarray mRNA expression data enhances computational identification of active microRNAs

Advanced computational predictive models of miRNA-mRNA interaction efficiency

The modeling of miRNA-mRNA interactions holds significant implications for synthetic biology and human health. However, this research area presents specific challenges due to the multifaceted nature of mRNA downregulation by miRNAs, influenced by numerous factors including competition or synergism among miRNAs and mRNAs. In this study, we present an improved computational model for predicting miRNA-mRNA interactions, addressing aspects not previously modeled. Firstly, we integrated a novel set of features that significantly enhanced the predictor's performance. Secondly, we demonstrated the cell-specific nature of certain aspects of miRNA-mRNA interactions, highlighting the importance of designing models tailored to specific cell types for improved accuracy. Moreover, we introduce a miRNA binding site interaction model (miBSIM) that, for the first time, accounts for both the distribution of miRNA binding sites along the mRNA and their respective strengths in regulating mRNA stability. Our analysis suggests that distant miRNA sites often compete with each other, revealing the intricate interplay of binding site interactions. Overall, our new predictive model shows a significant improvement of up to 6.43% over previous models in the field. The code of our model is available at https://www.cs.tau.ac.il/~tamirtul/miBSIM.

Predicting effective microRNA target sites in mammalian mRNAs

MicroRNA targets are often recognized through pairing between the miRNA seed region and complementary sites within target mRNAs, but not all of these canonical sites are equally effective, and both computational and in vivo UV-crosslinking approaches suggest that many mRNAs are targeted through non-canonical interactions. Here, we show that recently reported non-canonical sites do not mediate repression despite binding the miRNA, which indicates that the vast majority of functional sites are canonical. Accordingly, we developed an improved quantitative model of canonical targeting, using a compendium of experimental datasets that we pre-processed to minimize confounding biases. This model, which considers site type and another 14 features to predict the most effectively targeted mRNAs, performed significantly better than existing models and was as informative as the best high-throughput in vivo crosslinking approaches. It drives the latest version of TargetScan (v7.0; targetscan.org), thereby providing a valuable resource for placing miRNAs into gene-regulatory networks.

MixMir: microRNA motif discovery from gene expression data using mixed linear models

microRNAs (miRNAs) are a class of ∼22nt non-coding RNAs that potentially regulate over 60% of human protein-coding genes. miRNA activity is highly specific, differing between cell types, developmental stages and environmental conditions, so the identification of active miRNAs in a given sample is of great interest. Here we present a novel computational approach for analyzing both mRNA sequence and gene expression data, called MixMir. Our method corrects for 3' UTR background sequence similarity between transcripts, which is known to correlate with mRNA transcript abundance. We demonstrate that after accounting for kmer sequence similarities in 3' UTRs, a statistical linear model based on motif presence/absence can effectively discover active miRNAs in a sample. MixMir utilizes fast software implementations for solving mixed linear models, which are widely used in genome-wide association studies (GWASs). Essentially we use 3' UTR sequence similarity in place of population cryptic relatedness in the GWAS problem. Compared to similar methods such as miReduce, Sylamer and cWords, we found that MixMir performed better at discovering true miRNA motifs in three mouse Dicer-knockout experiments from different tissues, two of which were collected by our group. We confirmed these results on protein and mRNA expression data obtained from miRNA transfection experiments in human cell lines. MixMir can be freely downloaded from https://github.com/ldiao/MixMir.

Cell type-specific transcriptome analysis reveals a major role for Zeb1 and miR-200b in mouse inner ear morphogenesis

Cellular heterogeneity hinders the extraction of functionally significant results and inference of regulatory networks from wide-scale expression profiles of complex mammalian organs. The mammalian inner ear consists of the auditory and vestibular systems that are each composed of hair cells, supporting cells, neurons, mesenchymal cells, other epithelial cells, and blood vessels. We developed a novel protocol to sort auditory and vestibular tissues of newborn mouse inner ears into their major cellular components. Transcriptome profiling of the sorted cells identified cell type-specific expression clusters. Computational analysis detected transcription factors and microRNAs that play key roles in determining cell identity in the inner ear. Specifically, our analysis revealed the role of the Zeb1/miR-200b pathway in establishing epithelial and mesenchymal identity in the inner ear. Furthermore, we detected a misregulation of the ZEB1 pathway in the inner ear of Twirler mice, which manifest, among other phenotypes, malformations of the auditory and vestibular labyrinth. The association of misregulation of the ZEB1/miR-200b pathway with auditory and vestibular defects in the Twirler mutant mice uncovers a novel mechanism underlying deafness and balance disorders. Our approach can be employed to decipher additional complex regulatory networks underlying other hearing and balance mouse mutants.

Experimental design, preprocessing, normalization and differential expression analysis of small RNA sequencing experiments

Prior to the advent of new, deep sequencing methods, small RNA (sRNA) discovery was dependent on Sanger sequencing, which was time-consuming and limited knowledge to only the most abundant sRNA. The innovation of large-scale, next-generation sequencing has exponentially increased knowledge of the biology, diversity and abundance of sRNA populations. In this review, we discuss issues involved in the design of sRNA sequencing experiments, including choosing a sequencing platform, inherent biases that affect sRNA measurements and replication. We outline the steps involved in preprocessing sRNA sequencing data and review both the principles behind and the current options for normalization. Finally, we discuss differential expression analysis in the absence and presence of biological replicates. While our focus is on sRNA sequencing experiments, many of the principles discussed are applicable to the sequencing of other RNA populations.

Signatures of RNA binding proteins globally coupled to effective microRNA target sites

MicroRNAs (miRNAs) and small interfering RNAs (siRNAs), bound to Argonaute proteins (RISC), destabilize mRNAs through base-pairing with the mRNA. However, the gene expression changes after perturbations of these small RNAs are only partially explained by predicted miRNA/siRNA targeting. Targeting may be modulated by other mRNA sequence elements such as binding sites for the hundreds of RNA binding proteins (RNA-BPs) expressed in any cell, and this aspect has not been systematically explored. Across a panel of published experiments, we systematically investigated to what extent sequence motifs in 3' untranslated regions (UTRs) correlate with expression changes following transfection of small RNAs. The most significantly overrepresented motifs in down-regulated mRNAs are two novel U-rich motifs (URMs), UUUUAAA and UUUGUUU, recently discovered as binding sites for the ELAVL4 (also known as HuD) RNA-BP. Surprisingly, the most significantly overrepresented motif in up-regulated mRNAs is the heptanucleotide AU-rich element (ARE), UAUUUAU, which is known to affect mRNA stability via at least 20 different RNA-BPs. We show that destabilization mediated by the transfected miRNA is generally attenuated by ARE motifs and augmented by URM motifs. These ARE and URM signatures were confirmed in different types of published experiments covering eight different cell lines. Finally, we show that both ARE and URM motifs couple to presumed endogenous miRNA binding sites in mRNAs bound by Argonaute proteins. This is the first systematic investigation of 3' UTR motifs that globally couple to regulation by miRNAs and may potentially antagonize or cooperate with miRNA/siRNA regulation. Our results suggest that binding sites of miRNAs and RNA-BPs should be considered in combination when interpreting and predicting miRNA regulation in vivo.

The DIANA-mirExTra web server: from gene expression data to microRNA function

Background

High-throughput gene expression experiments are widely used to identify the role of genes involved in biological conditions of interest. MicroRNAs (miRNA) are regulatory molecules that have been functionally associated with several developmental programs and their deregulation with diverse diseases including cancer.

Methodology/Principal Findings

Although miRNA expression levels may not be routinely measured in high-throughput experiments, a possible involvement of miRNAs in the deregulation of gene expression can be computationally predicted and quantified through analysis of overrepresented motifs in the deregulated genes 3′ untranslated region (3′UTR) sequences. Here, we introduce a user-friendly web-server, DIANA-mirExTra (www.microrna.gr/mirextra) that allows the comparison of frequencies of miRNA associated motifs between sets of genes that can lead to the identification of miRNAs responsible for the deregulation of large numbers of genes. To this end, we have investigated different approaches and measures, and have practically implemented them on experimental data.

Conclusions/Significance

On several datasets of miRNA overexpression and repression experiments, our proposed approaches have successfully identified the deregulated miRNA. Beyond the prediction of miRNAs responsible for the deregulation of transcripts, the web-server provides extensive links to DIANA-mirPath, a functional analysis tool incorporating miRNA targets in biological pathways. Additionally, in case information about miRNA expression changes is provided, the results can be filtered to display the analysis for miRNAs of interest only.

Relative contribution of sequence and structure features to the mRNA binding of Argonaute/EIF2C-miRNA complexes and the degradation of miRNA targets

How miRNAs recognize their target sites is a puzzle that many experimental and computational studies aimed to solve. Several features, such as perfect pairing of the miRNA seed, additional pairing in the 3' region of the miRNA, relative position in the 3' UTR, and the A/U content of the environment of the putative site, have been found to be relevant. Here we have used a large number of previously published data sets to assess the power that various sequence and structure features have in distinguishing between putative sites that do and those that do not appear to be functional. We found that although different data sets give widely different answers when it comes to ranking the relative importance of these features, the sites inferred from most transcriptomics experiments, as well as from comparative genomics, appear similar at this level. This suggests that miRNA target sites have been selected in evolution on their ability to trigger mRNA degradation. To understand at what step in the miRNA-induced response individual features play a role, we transfected human HEK293 cells with miRNAs and analyzed the association of Argonaute/EIF2C-miRNA complexes with target mRNAs and the degradation of these messages. We found that structural features of the target site are only important for Argonaute/EIF2C binding, while sequence features such as the A/U content of the 3' UTR are important for mRNA degradation.