MAGIA: from miRNA and genes expression data integrative analysis to microRNA–transcription factor mixed regulatory circuits (2012 update)

MAGIA (http://gencomp.bio.unipd.it/magia2) is an update, extension and evolution of the MAGIA web tool. It is dedicated to the integrated analysis of in silico target prediction, microRNA (miRNA) and gene expression data for the reconstruction of post-transcriptional regulatory networks. miRNAs are fundamental post-transcriptional regulators of several key biological and pathological processes. As miRNAs act prevalently through target degradation, their expression profiles are expected to be inversely correlated to those of the target genes. Low specificity of target prediction algorithms makes integration approaches an interesting solution for target prediction refinement. MAGIA performs this integrative approach supporting different association measures, multiple organisms and almost all target predictions algorithms. Nevertheless, miRNAs activity should be viewed as part of a more complex scenario where regulatory elements and their interactors generate a highly connected network and where gene expression profiles are the result of different levels of regulation. The updated MAGIA tries to dissect this complexity by reconstructing mixed regulatory circuits involving either miRNA or transcription factor (TF) as regulators. Two types of circuits are identified: (i) a TF that regulates both a miRNA and its target and (ii) a miRNA that regulates both a TF and its target. INTRODUCTION MicroRNAs (miRNAs) are fundamental post-transcriptional regulators of several biological processes, whose alterations have been demonstrated in a number of diseases and in almost all human cancers (1–5). miRNA imperfectly bind the 30-untranslated region (30-UTR) of their target mRNAs and may cause translation inhibition and/or mRNA cleavage and degradation (6,7). miRNAs can have multiple targets and a single protein-coding gene can be targeted by multiple miRNAs. In this perspective, the network of post-transcriptional regulatory relationships is expected to have a highly complex connectivity structure. In silico target identification is based on (i) sequence similarity search, possibly considering target site evolutionary conservation and (ii) thermodynamic stability. However, it is known that the results of target prediction algorithms are characterized by very low specificity (8). This is caused both by the limited comprehension of the molecular basis of miRNA-target pairing and by the context-dependency of post-transcriptional regulation due to the cooperative interactions of different miRNAs. The integration of target predictions with miRNA and gene expression profiles has been recently proposed to improve the detection of functional miRNA-target relationships (9,10). As miRNAs act prevalently through target degradation, expression profiles of miRNA and target genes/transcripts are expected to be inversely correlated. Although some studies have demonstrated that the regulation of single key targets may largely explain the function of a given miRNA in tumorigenesis or metastasis development (11–13), the mechanisms of RNA regulation *To whom correspondence should be addressed. Tel: +39 049 827 7401; Fax: +39 049 827 6159; Email: [email protected] Correspondence may also be addressed to Stefania Bortoluzzi. Tel: +39 049 827 6502; Fax: +39 049 827 6209; Email: [email protected] The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors. Nucleic Acids Research, 2012, 1–9 doi:10.1093/nar/gks460 The Author(s) 2012. Published by Oxford University Press. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/ by-nc/3.0), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Nucleic Acids Research Advance Access published May 21, 2012 by gest on July 6, 2015 http://narrdjournals.org/ D ow nladed from are emerging to be much more complex than previously expected. miRNA activity should be viewed as part of a complex system where regulatory elements and their interactors generate a highly connected network, which regulates gene expression on multiple levels. Unfortunately, from an experimental point of view, the cooperation of different miRNAs acting on more than one target and the interplay with other regulatory levels are still challenging to be proved. The gene regulation at the transcriptional level is, in general, quantitatively stronger than that occurring post-transcriptionally. Both levels impact on mRNA/ genes expression profiles. Indeed, the miRNAs expression level can be activated or repressed by transcription factors (TFs), whereas mRNAs encoding TFs can be silenced by miRNAs. miRNAs and TF can share targets. Thus, miRNAs and TFs can form feedback or feed-forward loops, cooperating to tune gene expression (14) and playing critical roles in various biological processes. In a seminal review, Inui et al. (15) describes the role of miRNAs in pathways as decision maker elements to discriminate real versus too weak or too transient signals. In this context, the association between miRNAs and target expression profiles is generally more complex than linear correlation. Although many researchers have attempted to understand the mechanism of miRNAs to regulate target genes and their roles in various diseases, the study of miRNA regulation by TFs (TF–miRNA regulation) has been relatively limited (16). Some attempts have been reported in order to computationally reconstruct regulatory circuits concerning miRNA–TFs and their joint targets (17–19). However, the identification of miRNA promoters is a challenging task. As promoter regions could be several thousands of base pairs upstream of the primary-miRNA, a correct way of identifying putative miRNA regulatory regions was not clearly established so far. To this aim, a few databases have been developed to store experimentally validated TF-miRNAs interactions (16,20). Although several web tools (21–27) were recently developed in order to enhance the functional insights of miRNA functions and to improve miRNA–mRNA interactions identification almost none of them face the complex problem of the reconstruction of the above-mentioned TF–miRNA–target circuits. Here, we present MAGIA (miRNA and genes integrated analysis 2012 update, freely available at http://gencomp.bio.unipd.it/magia2) an updated and fully redesigned release of MAGIA (21), a web application for the integrative analysis of in silico target prediction, miRNA and gene expression data. As in the previous release, MAGIA refines in silico target prediction through miRNA–target expression association measures. MAGIA extends the analysis, supporting multiple organisms (human, mouse, rat and drosophila), and using a greatly expanded list of target predictions algorithms. It allows the combination of multiple predictions and the selection of user-defined thresholds for each one. The MAGIA architecture was completely redesigned: both data and analyses results are stored in a user-specific environment keeping the data private. Furthermore, the updated MAGIA tries to dissect regulatory complexity reconstructing mixed regulatory circuits involving either human miRNA or TF as regulators (Figure 1A). In particular, two types of mixed regulatory circuits are identified: (i) a TF that regulates both a given miRNA and its target gene; (ii) a miRNA that regulates both a given TF and its regulated gene (Figure 1B). In order to help the user in the interpretation of the results, novel features are implemented into MAGIA. Among them (i) miRNA–target interactions experimentally validated [as reported in miRecords (28), TarBase (29) and mirTarBase (30) databases] are specifically marked; (ii) Functional enrichment of the gene network component can be performed directly from MAGIA using DAVID platform. WEB TOOL IMPLEMENTATION In Figure 2, MAGIA architecture comprises three main sections: (i) data upload, (ii) data analysis and methods setup and (iii) results visualization, browsing and linking to external knowledgebase and tools.