lfa

lfa models latent population structure in binomial genomic data using logistic factor analysis to estimate latent factors in the natural parameter space.

Key Features:

• Probabilistic modeling: Formulates and estimates probabilistic models of genotype variation without requiring an admixture interpretation.
• Logistic factor analysis: Directly models the logit transformation of the probabilities underlying observed genotypes and uses latent variables to capture population structure.
• PCA integration: Incorporates principal component analysis techniques to estimate general probabilistic models that include established admixture models as special cases.
• Computational efficiency: Employs computationally efficient algorithms applicable to large-scale genomic datasets such as the Human Genome Diversity Panel and the 1000 Genomes Project.
• SNP differentiation identification: Identifies single nucleotide polymorphisms (SNPs) that exhibit significant differentiation with respect to inferred population structure under minimal model assumptions.

Scientific Applications:

• Population genetics analyses: Infers complex population structure in genome-wide genotyping data without relying on explicit admixture interpretations.
• SNP differentiation studies: Detects SNPs differentiated among populations for analyses of genetic variation and evolutionary patterns across diverse ancestries.
• Large-scale dataset analysis: Applies to consortium-scale datasets such as the Human Genome Diversity Panel and the 1000 Genomes Project to characterize ancestry-related structure.

Methodology:

Formulates and estimates probabilistic models by modeling the logit-transformed probabilities underlying observed genotypes with latent variables, integrates PCA-based estimation techniques, and employs computationally efficient algorithms.