Parental genotype reconstruction: applications of haplotype relative risk to incomplete parental data

Intended to resolve the problem of constructing a matched population-based control sample, haplotype relative risk techniques frequently suffer from loss of power for late-onset diseases due to unavailability of parental genotypes that are required to form parent-offspring pairs. However, much of this missing information can be reconstructed using the genotypes of the affected individual's siblings. For some cases, an estimate of the marker-allele frequencies for controls is needed. We have developed a technique based upon this idea and call it PGR (Parental Genotype Reconstruction). PGR represents a combination of the internal controls used by haplotype relative risk and controls based upon estimated allele frequencies. Although the latter is just what haplotype relative risk was designed to avoid, PGR has the potential to substantially boost power to detect linkage disequilibrium, for a relatively small increase in type I error. This performance is reasonably robust with respect to errors in the estimated allele frequencies, as we demonstrate by numerical simulation. A PGR software package has been written in FORTRAN and is available from Dr. Martin's web site at http://electro.psi.vcu.edu/rmartin/.