These are applications try to predict target binding sites for a specific miRNA or gene. The differences among applications are based on the algorithm used and also on different filtering factors such as the consideration of the secondary structure of the target mRNA, base pairing probabilities or free energy of the formed miRNA-mRNA hybrid. A personal and technical assessment of all the described applications is included. This section does not include algorithms for miRNA target prediction in plants (See the Plants section).
Parameters contributing to the final score
|Complementarity and free energy||Algorithm for finding genomic targets for miRNAs. Can be run locally in your computer and it is specially useful when working with "exotic" genomes.||
PROS: good for prediction of sites with imperfect binding within the seed region. There is a version of the software that can be locally installed and it is useful for analyzing non-conventional genomes.
CONS: low precision, too many false positives
|Complementarity, seed match, pairing probability||MirWalk is an integrated miRNA portal which includes its own target prediction algorithm, but also cross-talk with other algorithms.||
PROS: excellent interface and good performance. Very customizable interface for advanced target prediction using different criteria.
CONS: tendency to overestimate the number of targets when appeared in the same mRNA.
|Complementary, pairing probability and genomic content||miSTAR uses a logistic regression and random forest models to predict miRNA targets.||
PROS: good and easy to use interface. Uses the influence of binding site genomic context to increase accuracy.
CONS: obtained results are very accurate, but some non-canonical targets are missing.
|Binding energy, complementarity of the seed and conservation among species||Classical algorithm for the predcition of mIRNA targets in several organisms. It can be searched locally but all the data can be also downloaded by local use.||
PROS: miRNAs with high species conservation are more favourable to give good targeting scores.
CONS: It does not predict non-canonical sites
|Seed match, 3' complementarity, local AU content and position contribution||Prediction of miRNA targets for mammal genomes, fish, fly and C. elegans. It is frequently updated and can be searched by gene symbol or by miRNA ID. It gives information about the conservation of different miRNA families in all the scanned genomes.||
PROS: many parameters are included in the targeting score which is typically correlated with the protein downregulation in wet-lab experiments.
CONS: sites with poor seed pairing are often omitted.
|Seed match type||Similar to TargetScan but also focused on the search of mIRNA targets within ORFs of vertebrate genomes.||
PROS: similar to TargetScan but focused in vetebrate genomes. Simple tool for search conserved sites with strong seed pairing.
CONS: It usually underestimates miRNAs with multiple target sites.
|Free binding energy and complementarity||It is a integrated portal for miRNA target prediction and analysis. The prediction module is based on Artificial Neural Networks. It may be used to search for target genes of annotated or user defined miRNA sequences.||
PROS: it gives a score probability for each target site. Great interface.
CONS: Some miRNAs with multiple target sites in the same 3'-UTR may be omited.
|Target site accesibility and binding energy||Target prediction tool that takes into consideration the secondary structure of the target mRNA. The interface is easily customizable and can be used to search for miRNA, genes and also for UTR sequences.||
PROS: the secondary structure of the target mRNA is considered for the predictions.
CONS: low efficiency compared with other algorithms. Very optimistic.
|Pattern recognition and folding energy of the miRNA-mRNA hybrid||Algorithm for miRNA target prediction that uses input sequences for miRNA and UTR. Whole genomic predictions are available for download and browsing.||
PROS: Allows to identify sites targeted by new miRNAs.
CONS: low efficiency when compared with other algorithms