Admixture in Mexico City: implications for admixture mapping of Type 2 diabetes genetic risk factors

Abstract

Admixture mapping is a recently developed method for identifying genetic risk factors involved in complex traits or diseases showing prevalence differences between major continental groups. Type 2 diabetes (T2D) is at least twice as prevalent in Native American populations as in populations of European ancestry, so admixture mapping is well suited to study the genetic basis of this complex disease. We have characterized the admixture proportions in a sample of 286 unrelated T2D patients and 275 controls from Mexico City and we discuss the implications of the results for admixture mapping studies. Admixture proportions were estimated using 69 autosomal ancestry-informative markers (AIMs). Maternal and paternal contributions were estimated from geographically informative mtDNA and Y-specific polymorphisms. The average proportions of Native American, European and, West African admixture were estimated as 65, 30, and 5%, respectively. The contributions of Native American ancestors to maternal and paternal lineages were estimated as 90 and 40%, respectively. In a logistic model with higher educational status as dependent variable, the odds ratio for higher educational status associated with an increase from 0 to 1 in European admixture proportions was 9.4 (95%, credible interval 3.8–22.6). This association of socioeconomic status with individual admixture proportion shows that genetic stratification in this population is paralleled, and possibly maintained, by socioeconomic stratification. The effective number of generations back to unadmixed ancestors was 6.7 (95% CI 5.7–8.0), from which we can estimate that genome-wide admixture mapping will require typing about 1,400 evenly distributed AIMs to localize genes underlying disease risk between populations of European and Native American ancestry. Sample sizes of about 2,000 cases will be required to detect any locus that contributes an ancestry risk ratio of at least 1.5.

Appendix: Simulation of continuous gene flow

To simulate continuous gene flow, we consider a single meiosis in which the paternally derived and maternally derived gametes have admixture proportions μ ₁, μ _2, and sum of the intensities of the arrival processes (arrival rates) τ1, τ2. The density of ancestry state transitions on a gamete with admixture proportion μ and arrival rate τ is 2 μ(1-μ)τ. The mean density of ancestry state transitions on the two gametes is therefore μ ₁(1-μ ₁)τ ₁ + μ ₂(1-μ ₂)τ ₂. The meiosis generates a new gamete with admixture proportion μ _new = (μ ₁ + μ ₂)τ2. The proportion of meiotic crossovers that generate new transitions of ancestry is μ ₁(1-μ ₂) + (1-μ ₁)μ ₂. The arrival rate in the new gamete is therefore

$$ \tau _{{{\text{new}}}} {\text{ = }}\frac{{\mu _{{\text{1}}} {\left( {1 - \mu _{{\text{1}}} } \right)}\tau _{{\text{1}}} + \mu _{{\text{2}}} {\left( {1 - \mu _{{\text{2}}} } \right)}\tau _{{\text{2}}} + \mu _{{\text{1}}} {\left( {1 - \mu _{{\text{2}}} } \right)} + {\left( {1 - \mu _{{\text{1}}} } \right)}\mu _{{\text{2}}} }} {{2\mu _{{{\text{new}}}} {\left( {1 - \mu _{{{\text{new}}}} } \right)}}} $$

to simulate gene flow with proportion p of immigrants in each generation into an indigenous population of M gametes, we simulate each generation of new gametes from M meioses based on randomly-sampled pairs of gametes from the indigenous population (which after the first generation includes admixed individuals), and pM meioses based on pairing pM gametes derived from unadmixed immigrants with pM randomly-sampled gametes from the indigenous population. The mean arrival rate is evaluated only over admixed gametes, as unadmixed gametes contribute no information about this parameter. An R script for these simulations is available at http://www.ucd.ie/genepi/software.