Robust inference of population structure from next-generation sequencing data with systematic differences in...

Area of Science:

• Population genetics
• Genetic epidemiology
• Bioinformatics

Background:

• Inferring population structure is crucial for population genetics and genetic epidemiology.
• Principal Components Analysis (PCA) is effective for array data but challenging for sequencing data due to genotype uncertainty and variable sequencing quality.
• Existing PCA methods for sequencing data can produce biased results when sequencing properties differ across sample groups.

Purpose of the Study:

• To develop a robust PCA approach for inferring population structure from sequencing data, even with varying sequencing quality.
• To address biases in PCA caused by differences in read depth and error rates across sequencing groups.
• To enable accurate population structure inference when combining data from different studies or using historical control samples.

Main Methods:

• Developed a novel approach using sequencing reads directly, bypassing genotype calling.
• Implemented a subsampling procedure to standardize read depth across sequencing groups.
• Introduced a read-flipping procedure to harmonize error rates between groups.
• Averaged results over subsamples and read flips to preserve information.

Main Results:

• Demonstrated the utility of the new approach using two datasets from the 1000 Genomes Project.
• Validated the method's effectiveness through simulation studies.
• The method successfully adjusts for sequencing quality differences, preventing spurious components in PCA.

Conclusions:

• The new method enables accurate population structure inference from diverse sequencing data.
• This approach mitigates biases arising from systematic differences in sequencing quality.
• The developed TASER-PC software is publicly available for broader application.