Single-marker and multi-marker mixed models for polygenic score analysis in family-based data

Genome-wide association studies have proven successful but they remain underpowered for detecting variants of weaker effect. Alternative methods propose to test for association by using an aggregate score that combines the effects of the most associated variants. The set of variants that are to be aggregated may come from either of two modeling approaches: single-marker or multi-marker. The goal of this paper is to evaluate this alternative strategy by using sets of single-nucleotide polymorphisms identified by the two modeling approaches in the simulated pedigree data set provided for the Genetic Analysis Workshop 18. We focused on quantitative traits association analysis of diastolic blood pressure and of Q1, which served to control the statistical significance of our results. We carried out all analyses with knowledge of the underlying simulation model. We found that the probability to replicate association with the aggregate score depends on the single-nucleotide polymorphism set size and, for smaller sets (≤100), on the modeling approach. Nonetheless, assessing the statistical significance of these results in this data set was challenging, likely because of linkage because we are analyzing pedigree data, and also because the genotypes were the same across the replicates. Further methods need to be developed to facilitate the application of this alternative strategy in pedigree data.