I'm wondering if imputation, specifically Beagle, needs a reference panel that matches the sample's ancestry group. For example, Beagle documentation suggests the 1000 Genomes Project phase 3 reference panel, which contains all ancestry groups. Is this acceptable to use to impute the genome of, say, an East Asian, African, and European?
Basically, yes, you will get better results if you use a reference panel which contains samples with similar ancestry to your targets. Imputation works by matching haplotypes in your target samples to those in a reference panel. If the haplotypes in the target don’t exist in the reference panel, the variants won’t be imputed accurately.
We demonstrated that using TOPMed sequencing data as the imputation reference panel improves genotype imputation quality in these populations[Hispanic/Latino and African], which subsequently enhanced gene-mapping power for complex traits. For rare variants with minor allele frequency (MAF) < 0.5%, we observed a 2.3- to 6.1-fold increase in the number of well-imputed variants, with 11-34% improvement in average imputation quality, compared to the state-of-the-art 1000 Genomes Project Phase 3 and Haplotype Reference Consortium reference panels.
From this paper.
in my experience you should use a publicly cited reference genome like GRCh37 or GRCh38