If multiple Affymetrix probe sets were present for the same gene, the maximum fold change variant was selected.
For this work, the maximum population frequency of a variant was taken to be its maximal reported frequency in any data source.
The maximum area of HP of each variant was presented to better show the extreme observations such as 3589 ha for variable pFRAMe in variant IV.
Results for two models are presented here, each assuming a total MAF of 5% for all causal variants in the region: (a) the maximum MAF of any individual causal variant is 0.5%; and (b) the maximum MAF of any individual causal variant is 2%.
Note that the relative power of imputation appears lower under trait association model (A), where the maximum MAF of any causal variant is lower than under model (B).
The panels correspond to two specific trait association models: (A) the maximum MAF of any individual causal variant is 1%, and the total MAF of all causal variants is 5%; and (B) the maximum MAF of any individual causal variant is 0.5%, and the total MAF of all causal variants is 2%.
The minimum and maximum numbers of sequences assigned per variant were 1 and 2,513 sequence(s) per variant in pool A, and 1 and 12,688 in pool B. Variants occurring only once and variants containing indels were then removed from both datasets.
Overall, our results suggest that a reference panel of 4,000 individuals offers noticeable gains in power over 500 individuals only when the spectrum of causal variants is loaded with rarer variants (i.e. in our simulations, when the maximum MAF of any individual causal variant is 0.5%, rather than 1%).
A maximum of two allele variants was obtained per individual, indicating that our primers were specific for a single locus.
The maximum expected number of variants was set as 2 or 3 according to the ploidy level of the samples and variants were calculated using either all the mapped reads, or only using the uniquely mapped reads.
