Optimizing hard cutoff quality control metrics for IBD segment inference and genotyping accuracy in low-coverage...

Area of Science:

• Genomics
• Bioinformatics
• Population Genetics

Background:

• Identity-by-descent (IBD) inference from low-pass whole-genome sequencing (LPWGS) is vital for genetic studies.
• Genotype uncertainty in LPWGS can lead to biased relatedness estimates and inaccurate IBD inference.
• Robust quality control (QC) methods are essential for reliable IBD estimation.

Purpose of the Study:

• To evaluate the performance of genotype posterior probability (GP) and Bayes factor (BF) thresholds for QC in LPWGS.
• To determine optimal thresholds for balancing false positive and false negative rates in IBD inference.
• To develop a unified QC strategy for diverse coverage levels in LPWGS data.

Main Methods:

• A comprehensive grid search was performed to test various GP and BF thresholds.
• Performance was assessed using false positive rate, false negative rate, and mean squared error (MSE) of total IBD sharing.
• The study analyzed data across a range of coverage levels, from low to moderate-to-high.

Main Results:

• At moderate-to-high coverage (> 0.10×), GP > 0.99 provided stable performance, with minimal benefit from additional BF filtering.
• At low coverage (≤ 0.10×), GP-only filtering increased false positive IBD inference.
• Integrating BF with GP mitigated false positives at low coverage; a BF threshold > 100 balanced performance and marker retention.

Conclusions:

• Increasing GP thresholds alone does not consistently enhance IBD inference in LPWGS.
• Combining BF with GP offers improved inference stability, particularly under low coverage conditions.
• A unified QC strategy using both GP and BF is practical for low-to-moderate coverage LPWGS data.