Evaluating Power and Type 1 Error in Large Pedigree Analyses of Binary Traits

Studying population isolates with large, complex pedigrees has many advantages for discovering genetic susceptibility loci; however, statistical analyses can be computationally challenging. Allelic association tests need to be corrected for relatedness among study participants, and linkage analyses require subdividing and simplifying the pedigree structures. We have extended GenomeSIMLA to simulate SNP data in complex pedigree structures based on an Amish pedigree to generate the same structure and distribution of sampled individuals. We evaluated type 1 error rates when no disease SNP was simulated and power when disease SNPs with recessive, additive, and dominant modes of inheritance and odds ratios of 1.1, 1.5, 2.0, and 5.0 were simulated. We generated subpedigrees with a maximum bit-size of 24 using PedCut and performed two-point and multipoint linkage using Merlin. We also ran MQLS on the subpedigrees and unified pedigree. We saw no inflation of type 1 error when running MQLS on either the whole pedigrees or the sub-pedigrees, and we saw low type 1 error for two-point and multipoint linkage. Power was reduced when running MQLS on the subpedigrees versus the whole pedigree, and power was low for two-point and multipoint linkage analyses of the subpedigrees. These data suggest that MQLS has appropriate type 1 error rates in our Amish pedigree structure, and while type 1 error does not seem to be affected when dividing the pedigree prior to linkage analysis, power to detect linkage is diminished when the pedigree is divided.