been published on the genotype imputation procedures described above [59], [60]. These procedures are implemented in programs such as MERLIN [55], [61] and MENDEL [62], [63], using one of the pedigree analysis algorithms such as the Lander-Green [64] or Elston-Stewart [65] algorithms, or Monte Carlo sampling [66], [67]. Merlin uses sparse trees to represent gene flow in pedigrees and is considered as one of the fastest packages among packages implementing the same algorithms such as Allegro [68] and Genehunter [69]. In this study, we use MERLIN to infer the dosage probabilities. The output of this program includes the most likely genotypes, the expected number of copies for the tested alleles (0, 1, or 2 with genotype observed), and the posterior probabilities (dosage probabilities) of the three alternative genotypes [70]. Because a large number of related individuals are included, this family-based genotype inference is expected to improve the power of association tests [59]. Furthermore, when a GWA scan follows a linkage study, only a proportion of individuals may need to be genotyped and the inferred genotypes can be useful for the next step in the association analysis.
