Figure 1.
Relationship between ancestry and the distribution of neutrophil count.
(A) Distribution of normally transformed absolute neutrophil count for the three classes of genotype at rs2814778. Individuals who are homozygous for the null allele have distinctly lower neutrophil count (−0.35±0.89 standard deviations compared with the mean) than individuals who are carriers for the functional allele (0.76±0.89). We were able to place constraints on the frequency of the high neutrophil count allele in (B) West Africans, and (C) European Americans by assuming that the observed distributions of neutrophil count for each ancestry class (which we marked in practice by the genotype at rs2814778) are a mixture of distributions specified by the underlying allele frequency. The results indicate a 99% probability that the frequency is <4.9% in Africans and also a 99% probability that the frequency is >95.2% in Europeans.
Table 1.
Comparison of phenotypic characteristics for the four sets of samples used in this study.
Table 2.
Effect of the chromosome 1 locus on white blood cell counts.
Table 3.
Stratification of samples into 12 groups for admixture mapping analysis.
Figure 2.
Admixture association defines a 451 kb region containing the risk allele.
The LOD score for admixture association to neutrophil count shows a peak of 363.1, and a 99% confidence interval of 155.957–156.408 Mb (the region where the LOD score is within 1.44 of its maximum). The known genes under the peak are obtained using a screenshot of the “Known Genes” track from the UCSC genome browser (http://genome.ucsc.edu).
Table 4.
Reduced neutrophil count is more associated to the Duffy null polymorphism than to ancestry.
Figure 3.
Fine mapping reveals rs2814778 as the only significant association.
(A) Results of case-control association analysis for 193 SNPs genotyped in 148 individuals with low neutrophil count (<2,100/mm3), which we compared with 74 controls with high neutrophil count (5,000–9,000/mm3). All samples were selected to have a confident estimate of all African ancestry at the chromosome 1 locus (>99% probability) based on ANCESTRYMAP analysis at flanking markers outside the admixture peak. (B) HapMap SNPs of >5% minor allele frequency are well captured by this genotyping. We find that 94% of West African SNPs and 96% of European American SNPs are correlated with r2>0.8 to one of the SNPs we genotyped.
Figure 4.
Ancestry analysis of BAC clones that are the source of the human genome reference sequence across the chromosome 1 locus.
The human genome reference sequence across the admixture peak is pieced together from 5 BAC clones, which turn out to be a mosaic of European and African ancestry. To determine ancestry, we examined the haplotype of the human genome reference sequence for 284 SNPs for which data are available from the International Haplotype Map Project, and then output the ratio of the number of perfect matches to the reference sequence haplotype in 120 European American to the number of perfect matches in 120 West African chromosomes (conservatively adding 1 to the counts). Values above 10 indicate strong (>10∶1) evidence for a European haplotype, and values below 0.1 indicate strong (<1∶10) evidence for an African haplotype. RPCI-11, the individual who is the source of ∼74% of the human genome reference sequence, has two clones of entirely African and two clones of entirely European ancestry at this locus, strongly indicating that the human genome reference sequence is primarily that of an African American. We included these BAC clones in our search for polymorphisms across the admixture peak that were consistent with being fixed in frequency between European and African populations.