GRAF-pop: A Fast Distance-Based Method To Infer Subject Ancestry from Multiple Genotype Datasets Without Principal...

Area of Science:

• Genetics
• Bioinformatics
• Population Genetics

Background:

• Accurate subject ancestry inference is crucial for genetic association studies to address population stratification.
• Large-scale genetic data from diverse studies present challenges due to missing or unharmonized population data and varying genotyping platforms.
• Existing ancestry prediction methods often fail with incomplete genotype data, limiting their utility for integrated analysis.

Purpose of the Study:

• To develop a novel ancestry prediction method, GRAF-pop, robust to missing genotypes.
• To enable accurate and efficient ancestry inference and visualization across diverse genetic studies.
• To provide a method with a natural geometric interpretation for predicted population structure.

Main Methods:

• Developed GRAF-pop, a novel ancestry prediction method utilizing geometric and probabilistic approaches.
• GRAF-pop is designed to be robust to missing genotype data.
• Compared GRAF-pop's performance against PCA-based methods (EIGENSTRAT, FastPCA, FlashPCA2) using dense and sparse genotype data.

Main Results:

• GRAF-pop achieves comparable quality and running time to PCA-based methods with dense genotypes.
• GRAF-pop significantly outperforms PCA-based methods when dealing with sparse or missing genotype data.
• GRAF-pop provides visualized ancestry predictions with a clear geometric interpretation.

Conclusions:

• GRAF-pop offers a robust and accurate solution for ancestry inference, particularly in the presence of missing genotype data.
• The method has been successfully integrated into the dbGaP quality control process since February 2018.
• GRAF-pop enhances data harmonization and quality control for large genetic repositories, improving population structure analysis.