LongStitch: high-quality genome assembly correction and scaffolding using long reads

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-quality de novo genome assemblies are crucial for studying diverse organisms.
• Long-read sequencing significantly improves genome assembly and scaffolding by spanning repetitive regions.
• Existing methods struggle with resolving complex genomic structures, necessitating advanced tools.

Purpose of the Study:

• To present LongStitch, a scalable pipeline for correcting and scaffolding draft genome assemblies using exclusively long reads.
• To introduce ntLink, a novel long-read scaffolder integrated into the LongStitch pipeline.
• To evaluate the performance of LongStitch in improving genome assembly contiguity and accuracy.

Main Methods:

• LongStitch pipeline integrates Tigmint-long for assembly correction and ntLink/ARKS-long for scaffolding.
• The pipeline adapts existing tools (Tigmint-long, ARKS-long) for long-read data.
• ntLink utilizes lightweight minimizer mappings for efficient contig joining.

Main Results:

• LongStitch improved assembly contiguity by 1.2-fold to 304.6-fold (NGA50 length) across various species (C. elegans, O. sativa, human).
• The pipeline generated more contiguous and correct assemblies compared to the state-of-the-art scaffolder LRScaf.
• LongStitch achieved these improvements in under five hours with less than 23 GB of RAM.

Conclusions:

• LongStitch is an effective and efficient tool for improving draft genome assemblies using long reads.
• The pipeline is expected to benefit a wide range of de novo genome assembly projects.
• The LongStitch pipeline is publicly available for use in genomics research.