Integration of disparate sources of information to predict miRNA-mRNA interactions

Abstract

miRNAs are small RNA molecules (9 22nt) that interact with their corresponding target mRNAs inhibiting the translation of the mRNA into proteins and cleaving the target mRNA. This second effect diminishes the overall expression of the target mRNA. Several miRNA-mRNA relationship databases have been deployed, most of them based on sequence complementarities. However, the number of false positives in these databases is large and they do not overlap completely. In many cases the researcher has not clue on which database is the most suitable for his/her needs. Usually, the union of the databases is used in order to avoid missing possible interactions. This approach has an important drawback: all the databases are considered to be equally reliable (and is not the case).
In some cases, it is also interesting to combine miRNA and mRNA expression with the sequence based-predictions to achieve more accurate relationships.
In this work, we present a method that estimates the reliability of each interaction that appear in the databases. This method aggregates the information of the different databases using a Bayes approach. Using this information, we have created a metadabase that ouperforms published databases in both extension and quality. This database can be combined with mRNA and miRNA expression to find specific interactions that appear in a experiment. To this end, we used weighted LASSO regression with non-positive constraints (Talasso).
We show that the suggested metadatabase provides a meaningful score and that the top-ranked interactions are more enriched in experimentally validated interactions than any other database. In addition, this score can be combined with Talasso to get interactions that actually appear in a particular experiment.
TaLasso is available as Matlab or R code. There is also a web-based tool for human miRNAs at this address.