TreeLD

TreeLD infers ancestral genealogies across genomic regions using a coalescent framework to detect linkage disequilibrium (LD) signals and localize causative disease mutations by integrating genotype data with phenotype association models.

Key Features:

• Coalescent framework: Employs a coalescent model to sample genealogies across sampled chromosomes without initial conditioning on phenotype data.
• Genealogy sampling (MCMC): Uses Markov chain Monte Carlo (MCMC) to sample genealogies and generate posterior distributions of trees across chromosomes.
• Phenotype likelihood estimation: Averages over sampled genealogies to estimate phenotype likelihoods under models that include mutation and penetrance at an unobserved disease locus.
• Detection of nonrandom clustering: Identifies nonrandom clustering of chromosomes on the phylogenetic tree by phenotype to construct significance tests or Bayesian posterior distributions for localizing causative variants.
• Quantitative phenotype accommodation: Naturally incorporates quantitative phenotypic data in association analyses in addition to binary trait analysis.

Scientific Applications:

• Association testing and causative variation estimation: Unifies significance testing for association and estimation of causative variant location in LD-based mapping studies.
• Application to Mendelian and complex traits: Applied to Mendelian traits such as CFTR-related cystic fibrosis and complex traits such as type 2 diabetes linked to calpain-10.

Methodology:

Implements a two-stage inference: MCMC sampling of genealogies independent of phenotype, followed by averaging over sampled genealogies to compute phenotype likelihoods under mutation and penetrance models at an unobserved disease locus.