In conclusion, despite the acknowledgement of the necessity to collect large scale genome-wide data across different ancestries to fulfil the potential use of PGS in the precision medicine era3, this goal remains difficult to achieve in the near future. Instead, trans-ancestry studies have been increasingly popular. They incorporate genotype data from different ancestries to boost statistic power with increasing sample sizes, which have the benefit to discover disease/trait-associated loci and fine-mapping causal variants associated with complex traits or diseases13,40–43. However, the structure of the reference population still remains to be thoroughly explored, such as whether some specific populations with certain sample sizes are mostly useful in trans-ancestry studies. Our model presents an opportunity for such study design using both the LD and allele frequency information in a population level. By performing trans-ancestry GWASs, we expect that the predictive ability would increase when the admixed LD structure and allele frequency of the discovery population is similar to the target population.
