Improving miRNA-mRNA interaction predictions

Abstract

Background: MicroRNAs are short RNA molecules that post-transcriptionally regulate gene expression. Today, microRNA target prediction remains challenging since very few have been experimentally validated and sequence-based predictions have large numbers of false positives. Furthermore, due to the different measuring rules used in each database of predicted interactions, the selection of the most reliable ones requires extensive knowledge about each algorithm.

Results: Here we propose two methods to measure the confidence of predicted interactions based on experimentally validated information. The output of the methods is a combined database where new scores and statistical confidences are re-assigned to each predicted interaction. The new scores allow the robust combination of several databases without the effect of low-performing algorithms dragging down good-performing ones. The combined databases obtained using both algorithms described in this paper outperform each of the existing predictive algorithms that were considered for the combination.

Conclusions: Our approaches are a useful way to integrate predicted interactions from different databases. They reduce the selection of interactions to a unique database based on an intuitive score and allow comparing databases between them.

Figure 1

ROC curve for all predictive algorithms as well as the two combined approaches. The area under the curve and the number of interactions are also included for every algorithm.

Figure 2

Precision curve for all predicting algorithms and the combination methods. The curve was corrected by subtracting the precision value that corresponds to random interactions. The Y axis shows the Precision and the × axis the interactions sorted by score in descending order.

Figure 3

Description of the WSP and LRS methods. In the WSP method, box a), a new score for each interaction in each database is calculated by weighting their original scores with their associated accumulated precision. To this aim, for each of the databases, the interactions are sorted and their corresponding accumulated precisions are calculated. The obtained precision values are considered to be reliable in case they are larger than the randomly expected precision of the database. In the LRS method, box b), each interaction in each database is re-scored by assigning its probability of being experimentally-validated. To this aim, for each database, the probability of each interaction of being experimentally-validated is calculated. The probabilities in different databases are then combined considering their possible dependencies.