GeneScissors: a comprehensive approach to detecting and correcting spurious transcriptome inference owing to RNA-seq...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• RNA sequencing (RNA-seq) offers high-resolution transcriptome analysis.
• Multiple sequence alignments in RNA-seq data challenge existing analysis tools, leading to false positives and negatives.
• Current pipelines like TopHat/Cufflinks incorrectly identify pseudogenes and unannotated transcripts.

Purpose of the Study:

• To investigate genomic features causing multiple alignments and systematic errors in RNA-seq analysis.
• To develop a novel tool, GeneScissors, for detecting and correcting spurious transcriptome inference.
• To enhance the accuracy and reliability of RNA-seq data analysis.

Main Methods:

• Developed GeneScissors, a tool integrating machine learning with biological knowledge.
• Analyzed genomic features contributing to multiple alignments and their impact on transcriptome inference.
• Validated GeneScissors using simulated and real RNA-seq data.

Main Results:

• GeneScissors achieved ~90% accuracy in predicting spurious transcriptome calls from misalignments in simulations.
• Outperformed TopHat/Cufflinks and MapSplice/Cufflinks pipelines in precision and F-measurement.
• On real data, GeneScissors reduced pseudogene reporting by 53.6% and identified 0.97% more expressed, annotated transcripts, while flagging >16.3% of unannotated transcripts as false positives.

Conclusions:

• GeneScissors effectively corrects systematic errors in RNA-seq analysis stemming from multiple alignments.
• The tool significantly improves the precision of transcriptome inference, reducing false discoveries.
• GeneScissors enhances the identification of genuine expressed genes, contributing to more reliable downstream analyses.