MiRNAs

Predicting miRNA targets utilizing an extended profile HMM

Jan Grau, Daniel Arend, Ivo Grosse, Artemis G. Hatzigeorgiou, Jens Keilwagen, Manolis Maragkakis, Claus Weinholdt, and Stefan Posch

Abstract

The regulation of many cellular processes is influenced by miRNAs, and bioinformatics approaches for predicting miRNA targets evolve rapidly. Here, we propose conditional profile HMMs that learn rules of miRNA-target site interaction automatically from data. We demonstrate that conditional profile HMMs detect the rules implemented into existing approaches from their predictions. And we show that a simple UTR model utilizing conditional profile HMMs predicts target genes of miRNAs with a precision that is competitive compared to leading approaches, although it does not exploit cross-species conservation.

Graphical representation of learned models

Description of graphical representation: The darkness of the background of nodes represents how frequently nodes are visited according to the forward variables. The thickness of edges represents transition probabilities. The colored row vectors within the diamonds representing insert states reflect emission probabilities: green corresponds to A, blue corresponds to C, orange corresponds to G, and red corresponds to U. The saturation of these colors represents the corresponding emission probabilities and the brightness represents the deviation from a uniform distribution. In colored matrices within the rectangles representing match states, each row corresponds to one nucleotide observed in the miRNA in the order A, C, G, U and the colors within this row represent the conditional emission probabilities given the nucleotide in the miRNA in analogy to the emission probabilities of the insert states.

• Model learned on miRanda predictions
• Model learned on TargetScan predictions
• Model learned on RNAhybrid predictions
• Model learned on verified targets from mirecords
• Model used for predicting targets