2passtools

2passtools improves the accuracy of intron detection in long-read RNA sequencing by applying a two-pass alignment strategy that filters splice junctions using machine-learning-derived sequence information and alignment metrics.

Key Features:

• Two-pass alignment strategy: Performs an initial alignment pass followed by a guided realignment using filtered splice junctions.
• First-pass alignment to reference genome: Maps long reads to a reference genome to identify candidate splice junctions despite high long-read error rates.
• Machine-learning-derived sequence information: Incorporates machine-learning-derived sequence features to help distinguish genuine splice junctions from artifacts.
• Alignment metric filtering: Uses alignment metrics to filter spurious or erroneous splice junctions identified in the first pass.
• High-confidence splice junction generation: Produces a set of filtered, high-confidence splice junctions for use as alignment guides.
• Improved spliced alignments and transcriptome assembly: Refines spliced alignments and enhances the quality of transcriptome assemblies derived from long-read data.
• Applicability to unannotated species: Effective for organisms lacking high-quality genome annotations.
• Designed for long-read RNA sequencing: Explicitly addresses challenges from the relatively high error rates of long-read sequencing technologies.

Scientific Applications:

• Intron detection: Improves detection of introns in long-read RNA-seq datasets.
• Splice junction identification: Identifies and filters splice junctions to reduce alignment artifacts.
• Transcriptome assembly: Enhances the accuracy and quality of transcriptome assemblies from spliced long-read alignments.
• Alternative RNA processing analysis: Facilitates analysis of alternative splicing and other RNA processing events in eukaryotes.
• Cross-species splicing analysis: Enables splicing and transcriptome analyses in species without comprehensive annotations.

Methodology:

Perform an initial alignment of long reads to a reference genome, apply machine-learning-derived sequence information and alignment metrics to filter spurious splice junctions, then realign reads using the filtered high-confidence splice junctions as guides.