RT Journal Article SR Electronic T1 A Nested 2-Level Cross-validation Ensemble Learning Pipeline Suggests a Negative Pressure Against Crosstalk snoRNA-mRNA Interactions in Saccharomyces Cerevisae JF bioRxiv FD Cold Spring Harbor Laboratory SP 293555 DO 10.1101/293555 A1 Antoine Soulé A1 Jean-Marc Steyaert A1 Jéerôme Waldispuühl YR 2018 UL http://biorxiv.org/content/early/2018/04/02/293555.abstract AB The growing number of RNA-mediated regulation mechanisms identified in the last decades suggests a widespread impact of RNA-RNA interactions. The efficiency of the regulation relies on highly specific and coordinated interactions, while simultaneously repressing the formation of opportunistic complexes. However, the analysis of RNA interactomes is highly challenging due to the large number of potential partners, discrepancy of the size of RNA families, and the inherent noise in interaction predictions.We designed a recursive 2-step cross-validation pipeline to capture the specificity of ncRNA-mRNA interactomes. Our method has been designed to detect significant loss or gain of specificity between ncRNA-mRNA interaction profiles. Applied to snoRNA-mRNA in Saccharomyces Cerevisae, our results suggest the existence of a repression of ncRNA affinities with mRNAs, and thus the existence of an evolutionary pressure inhibiting such interactions.