Study of large and highly stratified population datasets by combining iterative pruning principal component analysis...

Area of Science:

• Population Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Large population genetic datasets challenge traditional structure analysis.
• Tracy-Widom (TW) test, widely used, shows susceptibility to type I errors in large datasets.
• Current Principal Component Analysis (PCA) methods resolve structure but not ancestry; model-based methods are unsuitable for large datasets.

Purpose of the Study:

• Introduce a novel framework for population structure analysis in large datasets.
• Develop a new heuristic, EigenDev, for robust structure detection.
• Integrate EigenDev with PCA methods to improve ancestry inference.

Main Methods:

• Developed and tested the EigenDev heuristic on simulated and real genetic data.
• Applied EigenDev to the iterative pruning PCA (ipPCA) method.
• Utilized ipPCA-derived subpopulation information to supervise STRUCTURE analysis for ancestry inference.

Main Results:

• EigenDev demonstrated robustness to sample size and outperformed the TW statistic, which exhibited type I errors in large samples.
• EigenDev integrated with ipPCA improved subpopulation number estimation and individual assignment accuracy.
• Analysis of bovine and human datasets revealed novel ancestry patterns consistent with ipPCA findings.

Conclusions:

• EigenDev is a superior heuristic for detecting population structure in large, complex datasets due to its sampling robustness.
• The EigenDev-ipPCA approach enhances subpopulation estimation and ancestry assignment accuracy.
• This framework provides a more comprehensive understanding of population structure by complementing parametric analyses.