Although haplotypes can provide great insight into the complex relationships between functional polymorphisms at a locus, their use in modern association studies has been limited. This is due to our inability to directly observe haplotypes in studies of unrelated individuals, but also to the extra complexity involved in their analysis and the difficulty in identifying which is the truly informative haplotype. Using a series of simulations, we tested a number of different models of a haplotype carrying two functional single nucleotide polymorphisms (SNPs) to assess the ability of haplotypic analysis to identify functional interactions between SNPs at the same locus. We found that, when phase is known, analysis of the haplotype is more powerful than analysis of the individual SNPs. The difference between the two approaches becomes less either as an increasing number of non-informative SNPs are included, or when the haplotypic phase is unknown, while in both cases the SNP association becomes progressively better at identifying the association. Our results suggest that when novel genotyping and bioinformatics methods are available to reconstruct haplotypic phase, this will permit the emergence of a new wave of haplotypic analysis able to consider interactions between SNPs with increased statistical power.

Although haplotypes can provide great insight into the complex relationships between functional polymorphisms at a locus, their use in modern association studies has been limited. This is due to our inability to directly observe haplotypes in studies of unrelated individuals, but also to the extra complexity involved in their analysis and the difficulty in identifying which is the truly informative haplotype. Using a series of simulations, we tested a number of different models of a haplotype carrying two functional single nucleotide polymorphisms (SNPs) to assess the ability of haplotypic analysis to identify functional interactions between SNPs at the same locus. We found that, when phase is known, analysis of the haplotype is more powerful than analysis of the individual SNPs. The difference between the two approaches becomes less either as an increasing number of non-informative SNPs are included, or when the haplotypic phase is unknown, while in both cases the SNP association becomes progressively better at identifying the association. Our results suggest that when novel genotyping and bioinformatics methods are available to reconstruct haplotypic phase, this will permit the emergence of a new wave of haplotypic analysis able to consider interactions between SNPs with increased statistical power.