We examined patterns of linkage disequilibrium (LD) among polymorphisms in drug-metabolizing enzyme (DME) genes. Polymorphisms within DME genes have been shown to alter drug responses in individuals. We have developed and wet-lab validated thousands of specific and robust TaqMan® drug metabolism genotyping assays for putatively functional polymorphisms (mostly SNPs, but also including multiple nucleotide and insertion/deletions variants) in 203 DME genes. This process allowed us to overcome the high degree of homology in these genes that challenges assay development. These variants were extensively annotated including names from locus-specific allele nomenclature committees, haplotypes, and public SNP identifiers. We genotyped 180 DNA samples from European, African-American, Chinese and Japanese populations with 1840 DME SNPs. Minor allele frequencies (MAF) in these variants were low (e.g. 0.17 in Europeans when polymorphic), as expected for alleles likely to be under strong purifying or directional selection. It has previously been noted that rare alleles are less likely to be in high LD with a set of ‘tagging SNPs’, such as those selected using HapMap data. We thus investigated the degree of pairwise LD among the DME SNPs. Within a gene, the distance between adjacent SNPs varied from 211 kb to 1 bp (total span 620 kb to 21 bp). Missense and especially nonsense SNPs were more likely to appear monomorphic (or have lower MAF than this study can detect) than the other types of variants (e.g. 3’ and 5’ UTR, splice site variants, and silent mutations): 27% of the 755 missense SNPs, 3% of the 64 nonsense SNPs, and 41% of the 1021 other variants were polymorphic in Europeans. Among polymorphic variants, coding SNPs had lower frequency (average MAF for the 202 missense SNPs was 0.14 and for the 2 nonsense SNPs was 0.06, compared to 0.19 for other variants). Few SNPs are in high or perfect linkage disequilibrium with each other: among 300 SNP pairs within 500 kb on chromosome 1, only 2.4% of the pairs have r2 > 0.85. The low frequency variants (MAF < 0.05) showed less LD with the other polymorphisms; e.g. on chromosome 1, the average pairwise r2 of low frequency variants was 0.037 while the average pairwise r2 of the higher frequency variants was 0.042. As expected, there was less LD between the SNPs in the African American population than in the other populations. Our results suggest that investigators interested in DME SNPs, and particularly in putative functional polymorphisms, should use caution in relying solely on tagging SNPs for research on disease association or drug efficacy, toxicity, metabolism, and should consider typing directly the relevant variants in their studies.
[Proc Amer Assoc Cancer Res, Volume 47, 2006]