Fine-mapping accuracy can also be improved by leveraging differences in LD patterns across continental populations that have arisen due to differences in demographic events such as population bottlenecks (see Figure 3)67–70. Intuitively, the set of tag SNPs linked to a causal variant will vary across populations, so that aggregating evidence of association across populations will dilute signals from tag SNPs and strengthen signals from causal variants. A standard approach to combining information across multiple studies is to compute posterior probabilities of causality from fixed-effects meta-analysis results67,69,71,72. Alternately, posterior probabilities can be computed from results of random-effects trans-ethnic meta-analysis methods64,68. These approaches assume a single causal variant and thus do not require LD information from the underlying populations. More recent studies have introduced hierarchical probabilistic models that allow for multiple causal variants while incorporating LD information from population reference panels61. These studies assume that causal variants are shared across populations but allow for heterogeneity in effect sizes across populations, and can also incorporate functional annotation data to further increase fine-mapping accuracy61. In an analysis of rheumatoid arthritis summary association data in
