Admixture

Admixture estimates individual ancestry proportions from multi-locus single nucleotide polymorphism (SNP) genotype data using a model-based maximum likelihood framework for population genetic analyses.

Key Features:

• Model-Based Ancestry Estimation: Estimates ancestry proportions and ancestral allele frequencies from SNP genotype data using a likelihood model analogous to STRUCTURE.
• Fast Numerical Optimization: Implements a block relaxation scheme with sequential quadratic programming (SQP) for block updates to accelerate computation.
• Quasi-Newton Acceleration: Incorporates a quasi-Newton method to improve convergence speed of the optimization.
• Comparative Speed and Accuracy: Achieves computational speeds comparable to EIGENSTRAT (PCA-based methods) while providing direct admixture fraction estimates.
• Accuracy of Estimates: Produces maximum likelihood estimates of admixture coefficients and ancestral allele frequencies shown to be as accurate as STRUCTURE's Bayesian estimates in simulations and empirical analyses.

Scientific Applications:

• Population Stratification Correction: Used in genetic association studies to estimate ancestry and correct for population stratification that can confound association signals.
• Large-Scale Genomic Analyses: Applicable to large SNP datasets and modern genomic studies that utilize extensive marker sets due to its computational efficiency.

Methodology:

Maximum likelihood estimation under a model-based framework similar to STRUCTURE, optimized by a block relaxation scheme with sequential quadratic programming for block updates and accelerated via a quasi-Newton method.