The snoGloBe interaction predictor enables a broader study of box C/D snoRNA functions and mechanisms

Deschamps-Francoeur, Gabrielle; Couture, Sonia; Abou-Elela, Sherif; Scott, Michelle S.

Box C/D small nucleolar RNAs (snoRNAs) are a conserved class of noncoding RNA known to serve as guides for the site-specific 2’-O-ribose methylation of ribosomal RNAs and the U6 small nuclear RNA, through direct base pairing with the target. In recent years however, several examples of box C/D snoRNAs regulating different levels of gene expression including transcript stability and splicing have been reported. These regulatory interactions typically require direct binding of the target but do not always involve the guide region. Supporting these new box C/D snoRNA functions, high- throughput RNA-RNA interaction datasets detect many interactions between box C/D snoRNAs and messenger RNAs. To facilitate the study of box C/D snoRNA functionality, we created snoGloBe, a box C/D snoRNA machine learning target predictor based on a gradient boosting classifier and considering snoRNA and target sequence and position as well as target type. SnoGloBe convincingly outperforms general RNA duplex predictors and PLEXY, the only box C/D snoRNA-specific target predictor available. The study of snoGloBe human transcriptome-wide predictions identifies enrichment in snoRNA interactions in exons and on exon-intron junctions. Some specific snoRNAs are predicted to target groups of functionally-related transcripts on common regulatory elements and the exact position of the predicted targets strongly overlaps binding sites of RNA-binding proteins involved in relevant molecular functions. SnoGloBe was also applied to predicting interactions between human box C/D snoRNAs and the SARS-CoV-2 transcriptome, identifying known and novel interactions. Overall, snoGloBe is a timely new tool that will accelerate our understanding of C/D snoRNA targets and function.