Mapping-free variant calling using haplotype reconstruction from k-mer frequencies

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Standard DNA variation detection relies on mapping short reads to a reference genome.
• This approach is limited for samples with dense variants or distant references.
• De novo assembly and hybrid methods are computationally expensive.

Purpose of the Study:

• To develop a novel algorithm for characterizing dense genetic variations.
• To create a software implementation (Kestrel) that bypasses traditional mapping and assembly.

Main Methods:

• Development of a novel k-mer algorithm named Kestrel.
• Implementation of Kestrel in Java for efficient computation.
• Application of Kestrel to identify single nucleotide polymorphisms (SNPs) and insertions/deletions (indels).

Main Results:

• Kestrel accurately characterized dense SNPs and large indels without mapping or assembly.
• Near-perfect concordance was observed when applied to mosaic penicillin binding protein (PBP) genes in Streptococcus pneumoniae.
• Multilocus sequence typing (MLST) was achieved without de novo assemblies, showing a low false-positive rate.
• Kestrel identified variants missed by other methods, though sensitivity limitations exist.

Conclusions:

• Kestrel offers a computationally efficient alternative for detecting complex genomic variations.
• The algorithm successfully bypasses the need for mapping or assembly in specific genomic contexts.
• Further development may address sensitivity limitations inherent in k-mer based approaches.