Admixed Populations: How To Correct For Ancestry Bias In Genetic Association Studies
Admixed populations pose a special challenge in genetic association studies due to the different proportions of a contribution of two or more ancestries. In the classic design of case-control studies, allele frequencies between disease and healthy subjects are compared; however, spurious associations not related to causative loci can be obtained by unmeasured population substructure. The issue of unobserved confounding effects appears when the investigated population is composed of several ancestral subpopulations with different allele frequencies and disease-risk not equally represented in cases and controls. This bias on allele frequencies can result in false-positive associations during statistical analyses.
The most common way to solve this is applying a set of ancestry informative markers (AIMs), i.e. markers exhibiting differential allele frequencies (d) greater than 30% in any pair of parental populations, to infer ancestry in both case and control groups and then adjusting the analysis for population stratification.
Using a whole genome-wide DNA microarray (Cytoscan HD array, Affymetrix) that encompasses 750,000 single nucleotide polymorphisms (SNPs), we reported 345 markers that are able to evaluate accurately the ancestral composition of the Brazilian population in case-control studies using data from the own array.
We performed a two-step validation of the 345 SNP-AIMs panel estimating the ancestral contributions using another panel of AIMs and ~70K SNPs from the array. The 345 SNP-AIMs panel has the potential to be widely used in cytogenetic research and molecular genetics to study diseases whose incidence is affected by ancestry. Another noteworthy advantage of using a panel of AIMs to infer ancestry is to insert the information of individual proportions of each parental component as an independent variable in statistical logistic regression models. This is especially suitable to perform corrections for ancestry in case-control studies.
In our study, we demonstrated the application of the panel comparing the ancestry in a case group of SLE versus healthy controls. The ancestry estimation based on the set of 345 SNP-AIMs showed that both groups had major European ancestral contributions, followed by African and Amerindian. Significant differences in the European and African ancestries were detected among the two different groups, while patients and controls have shown a major European genomic contribution. SLE patients have a higher African contribution (22%) than healthy subjects (13%), while controls showed a European contribution 12% higher than SLE patients. This difference throws light upon the moderate population substructure detected between the case and control groups.
We also highlighted the divergence between the self-declared ancestry and the individual genetic background. Comparisons between genetic and self-declared ancestry in SLE patients showed at least 30% of non-declared ancestry background, including African/Amerindian in Whites and European/Amerindian in Blacks. This finding underscores how crucial it is to evaluate ancestry based on genetic markers in addition to using the perception of individuals with their phenotypic characteristics.
In brief, spurious associations resulting from population stratification may be circumvented by using the 345 SNP-AIMs panel as a straightforward and efficient method to infer ancestry in case-control studies.
This study, Ancestry informative marker panel to estimate population stratification using genome-wide human array was recently published in the journal Annals of Human Genetics.