The rules of miRNA-target interactions

Description du projet

Identifier les cibles des microARN

Les microARN (miARN) sont de petites molécules d’ARN qui jouent un rôle essentiel dans la régulation de l’expression génétique. Des preuves de plus en plus nombreuses indiquent leur importance dans les contextes physiologiques et pathologiques, ce qui en fait un axe majeur de la recherche biomédicale. Cependant, la prédiction des cibles génétiques précises des miARN reste un défi informatique majeur. Le projet miRules, financé par l’UE, développera une stratégie bio-informatique avancée pour comprendre le mécanisme par lequel les miARN trouvent leurs cibles. Les chercheurs s’appuieront sur l’outil d’apprentissage profond miRBind et intégreront de nouvelles méthodologies pour améliorer la précision de la prédiction des cibles des miARN. Dans l’ensemble, le projet jettera les bases d’outils d’apprentissage automatique avancés qui soutiendront la recherche fondamentale et translationnelle dans le domaine de la biologie de l’ARN.

Objectif

Micro RNAs (miRNAs) are short regulatory RNA molecules which modulate the levels of RNA and proteins in cells.
Understanding the role of miRNAs in health and disease is a field with high engagement in molecular biology and biomedicine, with over 50 000 scientific articles published in the past 5 years (PubMed search “microrna[MeSH Terms]”).
However, predicting their molecular targets is still a major unsolved challenge, and the state-of-the-art bioinformatic solutions for this task remain far from perfect.
First, the task requires an exceptionally high accuracy and precision, as it can be compared to looking for a few hundred needles in a haystack: a few hundred targets among hundreds of thousands of transcripts. Second, the true mechanism of target binding is still unknown, which necessitates the use of heuristic approaches - ones that seem to work often enough, but are not necessarily related to the way miRNAs actually find their targets.

The aim of this project is to use bioinformatic methods to gain more knowledge about the mechanisms of miRNA-target interactions. We will approach this problem from several angles: 1) Using the deep learning methodology recently developed by the host group (miRBind), which reaches a single nucleotide resolution in its predictions; 2) Analyzing the interactions from an evolutionary perspective; 3) Developing a customized pairwise alignment algorithm combined with a dedicated statistical methodology. The Researcher will benefit from this project by gaining experience in the field of genomics in a multidisciplinary, collaborative environment.

In the future, the results of this project will enable creating knowledge-driven machine learning models for improved prediction of the molecular targets of miRNAs, which will facilitate the fundamental and applied research on the complex biology of these molecules.