Cost-effective sib-pair designs in the mapping of quantitative-trait loci

The extreme discordant-sib-pair design has been found to be the most powerful, across most genetic models. In this paper, we address two of the most frequently asked questions related to this design. First, under the extreme discordant-sib-pair design, a large number of people have to be screened for the phenotype of interest, before the desired number of discordant sibs can be collected for genotyping and linkage analysis. When the phenotyping cost is not negligible compared with the genotyping cost, such methods might not be cost effective. The second question is how sensitive the cost is to the genetic model and allele frequency. In this paper, we compare the cost under different sampling strategies, different genetic models, and different phenotyping:genotyping cost ratios. Because our knowledge of the underlying genetic model for a trait is limited, the discordant-sib-pair design proves to be the most robust. When the cost for screening probands is not included, the design that genotypes sibs with one sib in the top 10% and the other sib in the bottom 30% of the population with respect to the trait of interest is, across most models studied, the optimum among the designs considered in this paper. The cost under this design, across different genetic models, appears to be relatively robust to allele frequency and model type, whether additive or dominant. If probands initially must be screened as well, then 25% appears to be the optimal portions of the upper and lower distributions to be studied.