2passtools: two-pass alignment using machine-learning-filtered splice junctions increases the accuracy of intron...

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Eukaryotic transcription involves intricate alternative RNA processing.
• Long-read sequencing offers a comprehensive view of RNA processing complexity.
• High error rates in long-read data challenge accurate intron identification.

Purpose of the Study:

• To develop a computational method for enhancing the accuracy of intron identification from long-read RNA sequencing data.
• To improve transcriptome assembly by filtering spurious splice junctions.

Main Methods:

• Utilized alignment metrics and machine learning for filtering splice junctions.
• Implemented a two-pass realignment approach guided by validated junctions.
• Developed the 2passtools software package.

Main Results:

• Significantly improved the accuracy of spliced alignment for long-read RNA data.
• Enhanced the quality of transcriptome assembly for various species.
• The method is effective for both annotated and unannotated genomes.

Conclusions:

• The 2passtools method effectively filters spurious splice junctions, increasing accuracy.
• This approach advances the analysis of complex transcriptomes using long-read sequencing.
• Improved transcriptome assembly is achievable even with high-error sequencing data.