We now consider questions about possible differences in polygenic scores among ancestral populations. Polygenic scores, as currently calculated, vary with ancestry. Indeed, polygenic scoring practices from as early as 2009 accounted for this12. The method used by Purcell et al. in 200912 (and frequently since) includes two steps for mixed ancestry samples. First, samples are separated into more ancestrally homogeneous subgroups (using visual inspection of plots of principal components calculated on all genetic data from all samples). Second, principal components are calculated again within each of these more ancestrally homogeneous subgroups, and are used as covariates in polygenic scoring analyses, which are conducted separately within each subgroup. However, some research groups are not aware of these methodological recommendations and others (understandably) prefer a more inclusive analytical approach of analyzing all samples together (instead of creating subgroups). Figure 3 demonstrates why care must be taken in treatment of ancestry in polygenic scoring studies.
