LongStitch

LongStitch performs correction and scaffolding of draft de novo genome assemblies using long-read sequencing to improve assembly contiguity and resolve repetitive regions.

Key Features:

• Pipeline stages: Operates in up to three stages comprising initial misassembly correction followed by one or two incremental scaffolding stages for scalable processing of large datasets.
• Initial assembly correction (Tigmint-long): Uses Tigmint-long, adapted from linked-read methodologies, to identify and correct misassemblies in draft genomes.
• ntLink scaffolding: Employs ntLink, a long-read scaffolder that uses lightweight minimizer mappings of long reads to join contigs into scaffolds.
• ARKS-long scaffolding: Uses ARKS-long, an adaptation of linked-read ARKS, for additional long-read-based scaffolding and refinement.
• Contiguity improvement (NGA50): Demonstrates increases in NGA50 contiguity metrics of up to 304.6-fold on tested human assemblies.
• Comparative performance: Produces more contiguous and accurate assemblies than the tested long-read scaffolder LRScaf in comparative evaluations.
• Resource efficiency: Typical runs reported under five hours with peak memory usage below 23 GB RAM for evaluated datasets.
• Long-read support: Leverages long-read sequencing to span repetitive regions and resolve problematic assembly areas that challenge short-read assemblies.

Scientific Applications:

• De novo genome assembly improvement: Enhances draft assemblies for both model and non-model organisms by correcting misassemblies and increasing scaffold contiguity using long reads.
• Resolution of repetitive regions: Facilitates reconstruction of genomic regions containing repeats that are difficult to resolve with short-read data.
• Human genome assembly refinement: Applied to human genome assemblies to substantially increase contiguity metrics and enable comparative assembly assessment.

Methodology:

LongStitch applies Tigmint-long for initial misassembly correction, then performs successive scaffolding with ntLink (using lightweight minimizer mappings) and ARKS-long for further scaffolding and refinement.