EASTR: Identifying and eliminating systematic alignment errors in multi-exon genes

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Accurate RNA-to-genome alignment is crucial for gene expression analysis in biomedical research.
• Current splice-aware aligners like STAR and HISAT2 can produce erroneous spliced alignments, especially near repeated sequences.
• These errors can lead to 'phantom' introns and falsely spliced transcripts, impacting downstream analyses and even genome annotations.

Purpose of the Study:

• To address the issue of erroneous spliced alignments introduced by common RNA sequencing tools.
• To present EASTR (Emending Alignments of Spliced Transcript Reads), a novel software tool for detecting and removing false splice alignments and transcripts.
• To enhance the accuracy of RNA-seq data analysis and genome annotation.

Main Methods:

• Development of EASTR, a software tool that identifies sequence similarity between intron-flanking regions to detect erroneous alignments.
• Application of EASTR to alignment files generated by standard RNA sequencing workflows.
• Utilizing EASTR on existing genome annotation databases to identify and correct mis-annotations.

Main Results:

• EASTR effectively detects and removes falsely spliced alignments and transcripts caused by repeated sequences.
• Application of EASTR improves the accuracy of spliced alignments across various species, including human, maize, and Arabidopsis thaliana.
• Using EASTR before transcript assembly significantly reduces false positive introns, exons, and transcripts, leading to more accurate assembled transcripts.
• EASTR successfully identified and corrected mis-annotated transcripts within reference genome annotation databases.

Conclusions:

• EASTR is a valuable tool for improving the accuracy of RNA sequencing data analysis by correcting alignment errors.
• The software enhances the reliability of gene expression studies and the quality of genome annotations.
• EASTR's ability to detect sequence similarity provides a robust method for identifying artifactual splicing events.