TranscriptClean: variant-aware correction of indels, mismatches and splice junctions in long-read transcripts

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Long-read, single-molecule sequencing offers promise for discovering and characterizing RNA transcript isoforms.
• High error rates in these sequencing technologies hinder the accurate identification of novel transcript isoforms, often mistaking them for sequencing artifacts.

Purpose of the Study:

• To develop a computational tool, TranscriptClean, for correcting errors in long-read transcript sequencing data.
• To improve the reliability of isoform discovery and characterization from single-molecule sequencing platforms.

Main Methods:

• TranscriptClean was developed to correct mismatches, microindels, and noncanonical splice junctions in mapped transcripts.
• The package utilizes a reference genome to guide the correction process while preserving known biological variants.

Main Results:

• TranscriptClean effectively corrects nearly all mismatches and indels in human PacBio Iso-seq data.
• The method successfully rescues 39% of noncanonical splice junctions, significantly improving transcript data quality.

Conclusions:

• TranscriptClean is a valuable tool for enhancing the accuracy of long-read transcript isoform analysis.
• This package aids researchers in distinguishing genuine transcript isoforms from sequencing errors, advancing transcriptomics research.