RegScaf: a regression approach to scaffolding.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genome assembly correctness relies on accurate scaffolds, which order and orient contigs and define gap distances.
• Current scaffolding methods use linking read alignments, but these can be misled by factors like the contig boundary effect and repeats.
• Detecting incorrect linking information is a significant challenge in existing scaffolding approaches.

Purpose of the Study:

• To introduce RegScaf, a novel method for improving genome scaffolding accuracy.
• To address the limitations of current scaffolding techniques in handling erroneous read alignments and repeat regions.
• To enhance the precision of gap distance estimation in genome assemblies.

Main Methods:

• RegScaf analyzes contig distance distributions using kernel density estimation.
• It clusters orientation-supported links based on density modes to define linking distances.
• A linear model and least trimmed squares estimation are employed to robustly estimate contig positions and remove erroneous links.

Main Results:

• RegScaf demonstrates superior performance compared to popular scaffolders on synthetic and real datasets.
• The method significantly reduces abnormal errors, particularly in gap distance estimation.
• RegScaf effectively resolves repeat regions and shows adaptability to large genomes and long reads.

Conclusions:

• RegScaf offers a robust and accurate solution for genome scaffolding.
• The method's ability to mitigate errors from mistaken alignments improves overall genome assembly quality.
• RegScaf provides a valuable tool for genomic research, especially for complex genomes and long-read sequencing data.