Benchmarking datasets for assembly-based variant calling using high-fidelity long reads.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Long-read sequencing technologies have advanced genetic variant identification (variant calling).
• Benchmarking studies comparing different long-read sequencing platforms for variant calling are limited.

Purpose of the Study:

• To compare the variant calling performance of high-fidelity (HiFi) and continuous long-read (CLR) sequencing platforms.
• To evaluate different variant calling methods (read-based vs. assembly-based) across varying sequencing depths.

Main Methods:

• Utilized two Caenorhabditis elegans strains with known true-positive and induced variants.
• Generated both accurate and noisy long reads using HiFi and CLR sequencing.
• Compared variant calling metrics, including true-positive and false-positive rates, for both platforms and methods.

Main Results:

• HiFi sequencing identified 1.65-fold more true-positive variants and 60% fewer false-positive variants than CLR.
• Assembly-based variant calling was effective for detecting large insertions, even at 10× sequencing depth with accurate long reads.
• Variant calling after genome assembly demonstrated superior performance for accurate variant detection.

Conclusions:

• Variant calling after genome assembly using accurate long reads at ≥10× depth enables reliable true-positive variant detection.
• Propose 10× assembly-based variant calling as a cost-effective, high-quality methodology, especially considering HiFi sequencing costs.
• Findings may aid in developing population-level genetic variant identification methods.