Calling Structural Variants with Confidence from Short-Read Data in Wild Bird Populations

Area of Science:

• Population genomics
• Genomic variation analysis
• Bioinformatics

Background:

• Characterizing structural variation in natural populations is a recent advancement.
• High-confidence structural variant callsets are essential for population genomic studies.
• Short-read sequencing data presents challenges in accurately identifying structural variants.

Purpose of the Study:

• To compare heuristic-based quality filtering and visual curation methods for structural variant calling.
• To assess the impact of manual curation on the accuracy of structural variant datasets.
• To establish a cost-effective strategy for generating high-confidence structural variant callsets from short-read data.

Main Methods:

• Genome-wide structural variation landscape was characterized across 33 Nordic house sparrows (Passer domesticus).
• Heuristic-based quality filtering and visual curation (Samplot/PlotCritic, Samplot-ML) were compared.
• Manual curation strategies, including lenient and stringent approaches, were evaluated.

Main Results:

• Manual curation significantly reduces false positive structural variants, with a lenient strategy removing up to 80%.
• Variants rejected by manual curation showed no correlation with expected population structure from SNPs.
• Variants passing curation reflected expected population structure, indicating higher confidence.
• A lenient manual curation strategy by a single curator aligned with more stringent multi-curator assessments.

Conclusions:

• Manual curation is vital for improving the reliability of structural variant callsets derived from short-read data.
• Combining heuristic filtering with rapid manual curation offers a time- and cost-effective approach for population genomic studies.
• This strategy enhances the proportion of high-confidence variants, crucial for downstream functional and population analyses.