metaMIR is a microRNA (miRNA) framework to predict interactions in human between miRNAs and clusters of genes. The user provides a set of genes to be targeted, and optionally genes not to be targeted. Taking data from a reference database of previously established predictive algorithms, metaMIR will return miRNA candidates predicted to co-regulate genes among those entered by analyzing all possible subset combinations.

• Jason A. Davis, Sita J. Saunders, Martin Mann, Rolf Backofen
  Combinatorial ensemble miRNA target prediction of co-regulation networks with non-prediction data
  Nucleic Acids Research, 2017, 45 (15), 8745–8757.
  
• Martin Raden, Syed M Ali, Omer S Alkhnbashi, Anke Busch, Fabrizio Costa, Jason A Davis, Florian Eggenhofer, Rick Gelhausen, Jens Georg, Steffen Heyne, Michael Hiller, Kousik Kundu, Robert Kleinkauf, Steffen C Lott, Mostafa M Mohamed, Alexander Mattheis, Milad Miladi, Andreas S Richter, Sebastian Will, Joachim Wolff, Patrick R Wright, and Rolf Backofen
  Freiburg RNA tools: a central online resource for RNA-focused research and teaching
  Nucleic Acids Research, 46(W1), W25-W29, 2018.
  

Results are computed with metaMIR version 1.1.0 using R 3.2.1

Enter list of gene symbols (human, capitalized) to be analyzed from the list of supported gene IDs. Minimum 5. To indicate genes *not* to be targeted by an identified miRNA, prefix the symbol with a minus sign (-). To ensure that a particular gene is among any co-regulated cluster that is returned, prefix the symbol with an asterisk (*). Note that not all gene symbols are valid, due to use of alternate synonyms, or because there is no prediction data available.

The parameter constraints are: The genes have to be from the allowed gene list (see parameter help page). One gene per line and each gene might be prefixed with one of '-', '+', or '*'. Gene number (no duplicates) has to exceed Minimum number of genes per cluster.
Defaults to ()

Provide the minimum number of genes to be analyzed simultaneously.

The parameter constraints are: Input value has to be parsable as Integer. The value must be greater than or equal to 5 and must be smaller than or equal to 25.
Defaults to (5)

Provide the maximum number of genes to be analyzed simultaneously. If the gene list is longer than this maximum, the list will be partitioned by nearest-neighbor clustering into subsets for analysis (these will be returned in the output results).

The parameter constraints are: Input value has to be parsable as Integer. The value must be greater than or equal to 5 and must be smaller than or equal to 25. The value must be greater than or equal to the value of Minimum number of genes per cluster.
Defaults to (15)

The threshold for scores to be returned. No combinations/miRNAs will be returned with a score lower than this threshold. Note that lower scores will return more results and a threshold less than 1.0 may increase noise in permitting random results to be returned.

The parameter constraints are: Input value has to be parsable as Double. The value must be greater than or equal to 0.75 and must be smaller than or equal to 5.
Defaults to (1.1)

The specifics of the output will depend on the nature of the analysis; in each case, however, only the top score per miRNA is returned.

In standard mode, gene symbols are entered as-is, or only with asterisks. The result table will display the miRNA, its final calculated score (FinScore), the combination of genes predicted to simultaneously be targeted by the corresponding miRNA (PosCombo), the calculated score specific to this combination (PosScore), the number of genes in the cluster (PosGroup), and, if clustering occurred during processing, which list gave rise to this combination (PosList).
FinScore and PosScore will be the same unless the gene list to be analyzed was longer than the maximum cluster size. In this case, multiple lists are generated, and if the miRNA appears in more than one list, the results are averaged to generate the FinScore. The combination returned corresponds to the maximum individual score (PosScore) for that miRNA.

In differential mode, genes are additionally identified with minus signs (-) to indicate genes not to be targeted. The genes without prefix (to be targeted/down-regulated by the miRNA by prediction) are indicated by the PosCombo, PosScore, PosGroup, and PosList as described above. PosAg gives the aggregate score generated in the case of clustering for the positive gene set, again averaging the results across multiple lists. The Neg_ columns, correspond to those genes marked to not be targeted. In this case, if the number of negative genes is shorter than the minimum cluster size, the scores for the negative analysis are manually calculated. Otherwise, as for positive scores, the combinatorial analysis is performed, additionally with clustering if the number of "negative" genes is sufficiently long. The FinScore column then corresponds to the sum of the aggregate positive and negative scores, that is a miRNA will be scored more highly if it both targets the positive genes, while also avoiding genes not desired to be targeted.

An example investigating potential miRNA-mediated cross regulation between two cellular processes: the Hippo/YAP signaling pathway and transcription factors involved in neural crest (NC) development. Specifically, co-regulation here is investigated between those factors promoting Hippo/YAP transcriptional co-activity and transcription factors involved in the neural plate border specification during NC development.

The Hippo/YAP genes are marked with asterisks, to ensure that any combination returned in the analysis will have at least one YAP-related component subject to simultaneous co-regulation with NC genes. Otherwise, the results could be overwhelmed by the much longer list of NC genes, returning combinations that contain no Hippo/YAP elements.

The example's result can be directly accessed here

An example of metaMIR in differential analysis mode. The transforming growth factor beta (TGFb) signalling pathway, important during neural development (including neural crest) is often dysregulated in the development or progression of cancers. In many contexts, up-regulation of the activating components of the pathway can be involved in the progression of disease. A potential therapeutic application of miRNAs in this context would be to search a miRNA that targets (down-regulates) these components. At the same time, given the multitudinous targets of miRNAs, one would not want to target tumour suppressor genes.

In this gene list, the TGFb componens are plain, indicating the search should be "positive" to find miRNAs that will target (down-regulate) these genes as in the standard analysis. Genes prefixed with a minus sign (-) are those tumour suppressors one does *not* want to be simultaneously targeted by any miRNA returned for a cluster in the positive search.

The example's result can be directly accessed here

As described, the Hippo/YAP pathway is involved in regulation of cell proliferation (specifically at the transcriptional co-activation, or nuclear component end) and the TGFb signalling pathway can be dysregulated (by overactivity) in many cancers (see example descriptions for core and differential analyses above). It is thus feasible that miRNAs may exist that simultaneously target members of both of these pathways, as both may encourage normal or pathological cell proliferation.

This gene list is a combination of the core and differential analyses. The core is set to the YAP-related members of the Hippo/YAP signalling pathway (marked with asterisks), the TGFb genes are as previously (to be targeted in any combination; any, all or none may be returned in the final list), and in a hypothetical therapeutic context, any returned miRNA is not to target tumour suppressors (prefixed with minus signs).

The example's result can be directly accessed here

An example of a standard analysis in metaMIR, where the prediction is conducted such that all genes are potential targets for down-regulation by the returned miRNAs.

The example's result can be directly accessed here