Positional bias in variant calls against draft reference assemblies.

Area of Science:

• Genomics
• Bioinformatics
• Population Genetics

Background:

• Whole genome resequencing (WGR) enables population genetic and genome-wide association studies in non-model species using draft reference genomes.
• Draft assemblies, though lower in quality, are crucial for extending these analyses beyond model organisms.
• Understanding biases in variant calling pipelines is essential for accurate genomic data interpretation.

Purpose of the Study:

• To identify and characterize a positional bias in variant calling against draft reference assemblies.
• To investigate the impact of assembly algorithms and sequence characteristics on variant calling accuracy.

Main Methods:

• Assessed variant frequency at different positions within contigs and scaffolds.
• Analyzed publicly available draft assemblies and simulated short-read data.
• Investigated the effect of k-mer length, read length, and repetitive sequences on variant calling bias.

Main Results:

• A significant positional bias was detected in variant calling, with excess variants near contig/scaffold ends.
• This bias is linked to k-mer and read lengths and is exacerbated by repetitive sequences.
• Simulations confirmed false positives arise from reads of spatially distant repeats, particularly in contig assemblies.

Conclusions:

• Draft genome assemblies from common assemblers exhibit a positional bias affecting variant calling in non-model species.
• The bias stems from the inherent handling of repetitive sequences by assembly algorithms.
• Filtering variants in repetitive regions is recommended to enhance data quality when high-quality assemblies are not feasible.