Allele detection using k-mer-based sequencing error profiles

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genotyping and haplotype inference rely on aligning sequencing reads to a reference genome.
• Sequencing errors in long reads can lead to misidentification of alleles at variant sites, complicating accurate variant detection.

Purpose of the Study:

• To develop a method for modeling sequencing errors to improve allele detection accuracy.
• To enhance the performance of genotype and haplotype inference, especially for challenging sequencing data.

Main Methods:

• Introduced k-merald, a method to build sequencing error models using non-variant genomic regions.
• Utilized these error models to differentiate sequencing errors from true alternative alleles in variant regions.

Main Results:

• K-merald significantly improved allele detection accuracy, leading to better genotyping performance.
• Achieved an 18% and 24% reduction in error rate for Oxford Nanopore and PacBio CLR sequencing reads, respectively.
• Demonstrated substantial improvement in low-coverage sequencing data, with a 9% decrease in genotyping error rate for 3x Oxford Nanopore data.

Conclusions:

• K-merald provides a robust approach to account for systematic sequencing errors.
• The method enhances the reliability of genotype and haplotype inference from various sequencing technologies and coverages.