Computational identification of signals predictive for nuclear RNA exosome degradation pathway targeting

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• The RNA exosome degrades cellular transcripts in mammalian cell nucleoplasm.
• Substrate specificity is determined by the nuclear exosome targeting (NEXT) complex and the poly(A) exosome targeting (PAXT) pathway.
• Previous research identified some distinguishing DNA/RNA elements but lacked comprehensive analysis of their predictive power.

Purpose of the Study:

• To develop machine learning models for predicting RNA exosome targets.
• To identify and rank genomic features that distinguish between NEXT and PAXT targeting pathways.
• To discover novel genomic features associated with exosome targeting.

Main Methods:

• Utilized extensive genomic data to train machine learning models.
• Developed predictive models to classify RNA exosome targets.
• Assessed the predictive power of various genomic features, including transcript end sites, GC content, splice sites, and RNA helicase DDX3X binding sites.

Main Results:

• Machine learning models successfully predicted exosome targets with high accuracy.
• Features near transcript end sites, particularly the absence of canonical 3' end processing, were highly predictive of NEXT targets.
• Promoter-proximal GC content and 5' splice sites distinguished NEXT but not PAXT targets.
• RNA helicase DDX3X binding sites were identified as a novel predictive feature for exosome targeting.

Conclusions:

• Nucleoplasmic RNA exosome targeting is largely predictable using genomic features.
• The study provides an unbiased approach to assess the predictive power of known and novel features.
• Identified key features that differentiate NEXT and PAXT pathways, advancing understanding of RNA degradation regulation.