Allele frequency variation of association signals for schizophrenia in latin american populations

Genome-wide association studies (GWAS) to date have included mainly individuals of European ancestry; this has consequences for the detection of population-specific association signals and applicability of polygenic risk scores. We worked with the 342 linkage-disequilibrium independent significant SNPs identified in the most recent GWAS for schizophrenia. For these variants, we obtained allele frequencies for European and Latin American populations from 1000 genomes and gnomAD. We performed pairwise allele frequency comparisons between non-Finnish Europeans and the Latino/Admixed American population from gnomAD, and between non-Finnish Europeans and each of the Latin American populations from 1000 genomes (Mexico, Puerto Rico, Perú, Colombia). In a similar pairwise manner, we then compared the allele frequencies from non-Finnish Europeans and each of the four Latin American populations from 1000 genomes with the allele frequencies from the Amerindigenous component of the Latino/Admixed population available in gnomAD (estimated through local ancestry inference). While the average difference in allele frequency between non-Finnish Europeans and admixed Americans is 6%, when the comparison is made with each individual Latin American population the average differences in allele frequency range from 5,5% (non-Finnish European vs Puerto Rico) to 14% (non-Finnish European vs Perú). Additionally, we observed that the average differences in allele frequencies between each Latin American population as a whole and the Amerindigenous component are 16% for Puerto Rico, 13% for Colombia, 10% for Mexico, and 7% for Perú. This reflects variations in admixture in different Latin American populations. Our results illustrate that the Latino/Admixed American (sometimes called “Hispanic”) category often used in cross-ancestry GWAS, not only presents the challenge of including three-way admixed individuals (African/European/Amerindigenous), but also groups together heterogeneous populations with differing allele frequencies and admixture composition. Inclusion of different Latin American populations in future GWAS, especially those with a large Amerindigenous component, can result in detection of association signals with new variants or with variants with very low frequencies in populations of European ancestry. A better representation of diverse populations (not only Latin American) in GWAS can improve our understanding of the biological mechanisms of disease, as well as the performance of polygenic risk scores in these populations.