Genotypes are typically represented as the discrete counts of the minor or effect allele (0, 1, or 2), for SNPs, in each individual. Genotypes not included in the genotyping chip can, potentially, be imputed and are usually either recorded as a set of 3 probabilities corresponding to the probability of each of the possible genotypes [19] or, based on these, as the expected genotype (a real number between 0 and 2 known as the “dosage”) [19] or as the “best-guess” (most probable) genotype. While any of these data formats can be exploited in PRS analyses, the most common approach is to use the best-guess genotype for each individual. However, this approach does not account for the uncertainty in the imputed genotype.
