Inferring Population Structure and Admixture Proportions in Low-Depth NGS Data

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Population structure and admixture inference are crucial for population genetics and association studies.
• Traditional methods like Principal Component Analysis (PCA) face challenges with low-depth next-generation sequencing (NGS) data due to statistical uncertainty.
• Accounting for uncertainty using genotype likelihoods is key for accurate analysis.

Purpose of the Study:

• To present two novel methods for inferring population structure and admixture proportions from low-depth NGS data.
• To enhance the accuracy of population genetics analyses with variable sequencing depths.
• To provide a robust computational framework for these analyses.

Main Methods:

• An iterative heuristic approach for Principal Component Analysis (PCA) estimating individual allele frequencies directly from genotype likelihoods.
• A fast non-negative matrix factorization method utilizing estimated individual allele frequencies for admixture proportion inference.
• Implementation of both methods within the PCAngsd software framework.

Main Results:

• Demonstrated improved accuracy in inferring population structure from low and variable depth sequencing data.
• Validated the methods using both simulated and real-world biological datasets.
• Showcased the efficiency of the non-negative matrix factorization approach for admixture analysis.

Conclusions:

• The proposed methods effectively handle statistical uncertainty in low-depth NGS data.
• These tools offer enhanced accuracy for population structure and admixture inference in genetic studies.
• The PCAngsd framework provides a valuable resource for population geneticists and bioinformaticians.