DoTT-ML: Condition-aware detection of transcriptional readthrough from RNA-seq with optional ML-based prioritization.

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• Disruption of transcription termination (DoTT) leads to aberrant readthrough RNA transcripts.
• Existing tools for DoTT detection have limitations in direct condition-to-condition analysis.
• DoTT is observed under various cellular stresses, including viral infections and metabolic changes.

Purpose of the Study:

• To develop a condition-aware computational pipeline, DoTT-ML, for detecting transcription termination disruption from RNA-seq data.
• To enable direct comparison of DoTT events between different experimental conditions.
• To validate the biological relevance of DoTT-ML findings in a metabolic disease model.

Main Methods:

• DoTT-ML extends gene annotations with a tunable downstream window and an optional gap to mitigate noise.
• It employs a robust statistical workflow for differential analysis between conditions.
• An optional machine learning component aids in prioritizing findings when reference annotations are available.

Main Results:

• DoTT-ML was benchmarked against existing tools (ARTDeco, DoGFinder) on diverse RNA-seq datasets, demonstrating comparable or superior performance (high ROC AUC).
• The pipeline identified diet-associated readthroughs in metabolic genes within a mouse high-carbohydrate diet model.
• Experimental validation confirmed a diet-inducible readthrough transcript at the Scd1 locus.

Conclusions:

• DoTT-ML offers a practical and effective framework for condition-aware detection and comparison of transcription termination disruption.
• The pipeline facilitates the discovery of biologically relevant readthrough events, particularly under stress conditions like metabolic perturbation.
• DoTT-ML enhances the analysis of diverse RNA-seq assays for studying gene regulation and cellular responses.