HAPPY

HAPPY estimates haplotype-specific odds ratios from unphased genotype data by using inferred (PHASEd) haplotypes as covariates in logistic regression to assess haplotype associations and haplotype×environment interactions in case-control studies.

Key Features:

• Haplotype inference: Infers individual haplotypes from unphased genotypic data (PHASEd haplotypes) and uses inferred haplotypes as covariates in epidemiologic regression analyses, including conditional logistic regression.
• Adjustment for uncertainty: Incorporates strategies to adjust for uncertainty in inferred haplotypes to account for phase ambiguity in association estimates.
• Analytic strategies comparison: Compares analytic strategies via simulation, including most-likely haplotype assignment, expectation substitution as described by Stram et al. (2003), and an improper multiple imputation approach in matched and unmatched case-control contexts.
• Regression framework: Implements unconditional logistic regression for unrelated cases and controls and adjusts main effects of covariates to estimate stratum-specific haplotype effects.
• Tests and performance metrics: Conducts omnibus and haplotype-specific tests of association and evaluates methods by bias and mean squared error for haplotype and haplotype×environment interaction estimates.

Scientific Applications:

• Association studies: Quantifies associations between haplotypes of single nucleotide polymorphisms (SNPs) and phenotypes or diseases in population-based case-control studies.
• Tagging SNPs: Supports analyses using tagging SNPs as surrogates for unobserved causal variation across genomic regions.
• Haplotype×environment interactions: Estimates both haplotype main effects and haplotype×environment interaction effects.
• Applied example: Has been applied to progesterone-receptor haplotypes and endometrial cancer to illustrate how haplotype tagging of causal variants affects analytic performance.

Methodology:

Inferring individual haplotypes from unphased genotypic data; implementing unconditional logistic regression with covariate adjustment; conducting omnibus and haplotype-specific association tests; evaluating strategies for handling uncertainty in inferred haplotypes via simulation including most-likely assignment, expectation substitution (Stram et al. 2003), and improper multiple imputation; assessing bias and mean squared error.