Michael A. Eberle

Selecting a Maximally Informative Set of Single-Nucleotide Polymorphisms for Association Analyses Using Linkage Disequilibrium

Common genetic polymorphisms may explain a portion of the heritable risk for common diseases. Within candidate genes, the number of common polymorphisms is finite, but direct assay of all existing common polymorphism is inefficient, because genotypes at many of these sites are strongly correlated. Thus, it is not necessary to assay all common variants if the patterns of allelic association between common variants can be described. We have developed an algorithm to select the maximally informative set of common single-nucleotide polymorphisms (tagSNPs) to assay in candidate-gene association studies, such that all known common polymorphisms either are directly assayed or exceed a threshold level of association with a tagSNP. The algorithm is based on the r(2) linkage disequilibrium (LD) statistic, because r(2) is directly related to statistical power to detect disease associations with unassayed sites. We show that, at a relatively stringent r(2) threshold (r2>0.8), the LD-selected tagSNPs resolve >80% of all haplotypes across a set of 100 candidate genes, regardless of recombination, and tag specific haplotypes and clades of related haplotypes in nonrecombinant regions. Thus, if the patterns of common variation are described for a candidate gene, analysis of the tagSNP set can comprehensively interrogate for main effects from common functional variation. We demonstrate that, although common variation tends to be shared between populations, tagSNPs should be selected separately for populations with different ancestries.

Mapping complex disease loci in whole-genome association studies

Identification of the genetic polymorphisms that contribute to susceptibility for common diseases such as type 2 diabetes and schizophrenia will aid in the development of diagnostics and therapeutics. Previous studies have focused on the technique of genetic linkage, but new technologies and experimental resources make whole-genome association studies more feasible. Association studies of this type have good prospects for dissecting the genetics of common disease, but they currently face a number of challenges, including problems with multiple testing and study design, definition of intermediate phenotypes and interaction between polymorphisms.

Population history and natural selection shape patterns of genetic variation in 132 genes.

Identifying regions of the human genome that have been targets of natural selection will provide important insights into human evolutionary history and may facilitate the identification of complex disease genes. Although the signature that natural selection imparts on DNA sequence variation is difficult to disentangle from the effects of neutral processes such as population demographic history, selective and demographic forces can be distinguished by analyzing multiple loci dispersed throughout the genome. We studied the molecular evolution of 132 genes by comprehensively resequencing them in 24 African-Americans and 23 European-Americans. We developed a rigorous computational approach for taking into account multiple hypothesis tests and demographic history and found that while many apparent selective events can instead be explained by demography, there is also strong evidence for positive or balancing selection at eight genes in the European-American population, but none in the African-American population. Our results suggest that the migration of modern humans out of Africa into new environments was accompanied by genetic adaptations to emergent selective forces. In addition, a region containing four contiguous genes on Chromosome 7 showed striking evidence of a recent selective sweep in European-Americans. More generally, our results have important implications for mapping genes underlying complex human diseases.

Additional SNPs and linkage-disequilibrium analyses are necessary for whole-genome association studies in humans

More than 5 million single-nucleotide polymorphisms (SNPs) with minor-allele frequency greater than 10% are expected to exist in the human genome. Some of these SNPs may be associated with risk of developing common diseases. To assess the power of currently available SNPs to detect such associations, we resequenced 50 genes in two ethnic samples and measured patterns of linkage disequilibrium between the subset of SNPs reported in dbSNP and the complete set of common SNPs. Our results suggest that using all 2.7 million SNPs currently in the database would detect nearly 80% of all common SNPs in European populations but only 50% of those common in the African American population and that efficient selection of a minimal subset of SNPs for use in association studies requires measurement of allele frequency and linkage disequilibrium relationships for all SNPs in dbSNP.

Haplotype Diversity across 100 Candidate Genes for Inflammation, Lipid Metabolism, and Blood Pressure Regulation in Two Populations

Recent studies have suggested that a significant fraction of the human genome is contained in blocks of strong linkage disequilibrium, ranging from ~5 to >100 kb in length, and that within these blocks a few common haplotypes may account for >90% of the observed haplotypes. Furthermore, previous studies have suggested that common haplotypes in candidate genes are generally shared across populations and represent the majority of chromosomes in each population. The conclusions drawn from these preliminary studies, however, are based on an incomplete knowledge of the variation in the regions examined. To bridge this gap in knowledge, we have completely resequenced 100 candidate genes in a population of African descent and one of European descent. Although these genes have been well studied because of their medical importance, we demonstrate that a large amount of sequence variation has not yet been described. We also report that the average number of inferred haplotypes per gene, when complete data is used, is higher than in previous reports and that the number and proportion of all haplotypes represented by common haplotypes per gene is variable. Furthermore, we demonstrate that haplotypes shared between the two populations constitute only a fraction of the total number of haplotypes observed and that these shared haplotypes represent fewer of the African-descent chromosomes than was expected from previous studies. Finally, we show that restricting variation discovery to coding regions does not adequately describe all common haplotypes or the true haplotype block structure observed when all common variation is used to infer haplotypes. These data, derived from complete knowledge of genetic variation in these genes, suggest that the haplotype architecture of candidate genes across the human genome is more complex than previously suggested, with important implications for candidate gene and genomewide association studies.

A new susceptibility locus for autosomal dominant pancreatic cancer maps to chromosome 4q32-34.

Pancreatic cancer is the fifth leading cause of cancer death in the United States. Nearly every person diagnosed with pancreatic cancer will die from it, usually in <6 mo. Familial clustering of pancreatic cancers is commonly recognized, with an autosomal dominant inheritance pattern in approximately 10% of all cases. However, the late age at disease onset and rapid demise of affected individuals markedly hamper collection of biological samples. We report a genetic linkage scan of family X with an autosomal dominant pancreatic cancer with early onset and high penetrance. For the study of this family, we have developed an endoscopic surveillance program that allows the early detection of cancer and its precursor, before family members have died of the disease. In a genomewide screening of 373 microsatellite markers, we found significant linkage (maximum LOD score 4.56 in two-point analysis and 5.36 in three-point analysis) on chromosome 4q32-34, providing evidence for a major locus for pancreatic cancer.

Lower-Than-Expected Linkage Disequilibrium between Tightly Linked Markers in Humans Suggests a Role for Gene Conversion

Understanding the pattern of linkage disequilibrium (LD) in the human genome is important both for successful implementation of disease-gene mapping approaches and for inferences about human demographic histories. Previous studies have examined LD between loci within single genes or confined genomic regions, which may not be representative of the genome; between loci separated by large distances, where little LD is seen; or in population groups that differ from one study to the next. We measured LD in a large set of locus pairs distributed throughout the genome, with loci within each pair separated by short distances (average 124 bp). Given current models of the history of the human population, nearly all pairs of loci at such short distances would be expected to show complete LD as a consequence of lack of recombination in the short interval. Contrary to this expectation, a significant fraction of pairs showed incomplete LD. A standard model of recombination applied to these data leads to an estimate of effective human population size of 110,000. This estimate is an order of magnitude higher than most estimates based on nucleotide diversity. The most likely explanation of this discrepancy is that gene conversion increases the apparent rate of recombination between nearby loci.

Whole-genome haplotyping by dilution, amplification, and sequencing

Standard whole-genome genotyping technologies are unable to determine haplotypes. Here we describe a method for rapid and cost-effective long-range haplotyping. Genomic DNA is diluted and distributed into multiple aliquots such that each aliquot receives a fraction of a haploid copy. The DNA template in each aliquot is amplified by multiple displacement amplification, converted into barcoded sequencing libraries using Nextera technology, and sequenced in multiplexed pools. To assess the performance of our method, we combined two male genomic DNA samples at equal ratios, resulting in a sample with diploid X chromosomes with known haplotypes. Pools of the multiplexed sequencing libraries were subjected to targeted pull-down of a 1-Mb contiguous region of the X-chromosome Duchenne muscular dystrophy gene. We were able to phase the Duchenne muscular dystrophy region into two contiguous haplotype blocks with a mean length of 494 kb. The haplotypes showed 99% agreement with the consensus base calls made by sequencing the individual DNAs. We subsequently used the strategy to haplotype two human genomes. Standard genomic sequencing to identify all heterozygous SNPs in the sample was combined with dilution-amplification–based sequencing data to resolve the phase of identified heterozygous SNPs. Using this procedure, we were able to phase >95% of the heterozygous SNPs from the diploid sequence data. The N50 for a Yoruba male DNA was 702 kb whereas the N50 for a European female DNA was 358 kb. Therefore, the strategy described here is suitable for haplotyping of a set of targeted regions as well as of the entire genome.

An analysis of strategies for discovery of single-nucleotide polymorphisms

Strategies for the discovery of single‐nucleotide polymorphisms (SNPs) can be characterized by the number of individuals in the discovery sample, and by the minimal required number of observations of each allele. We examined the effect of different strategies on two key properties of the resulting SNP collection: (1) the probability that a SNP with a given population allele frequency is detected; and (2) the allele‐frequency distribution of the discovered SNPs. We show that strategies that accept all polymorphic sites lead to collections with a high fraction of SNPs with rare minor alleles, particularly in expanded populations. Such SNPs have a low probability of replication in a second sample. We discuss how to tailor a discovery strategy to the desired properties of a SNP collection. Genet. Epidemiol. 19(Suppl 1):S29–S35, 2000.

Improved imputation of common and uncommon SNPs with a new reference set

6 volume 44 | number 1 | january 2012 | nature genetics Statistical imputation of genotype data is an important statistical technique that uses patterns of linkage disequilibrium observed in a reference set of haplotypes to computationally predict genetic variants in silico1. Currently, the most popular reference sets are the publicly available International HapMap2 and 1000 Genomes data sets3. Although these resources are valuable for imputing a sizeable fraction of common SNPs, they may not be optimal for imputing data for the next generation of genome-wide association studies (GWAS) and SNP arrays, which explore a fraction of uncommon variants. We have built a new resource for the imputation of SNPs for existing and future GWAS, known as the Division of Cancer Epidemiology and Genetics (DCEG) Reference Set. The data set has genotypes for cancer-free individuals, including 728 of European ancestry from three large prospectively sampled studies4–6, 98 AfricanAmerican individuals from the Prostate, Lung, Colon and Ovary Cancer Screening Trial (PLCO), 74 Chinese individuals from a clinical trial in Shanxi, China (SHNX)7 and 349 individuals from the HapMap Project (Table 1). The final harmonized data set includes 2.8 million autosomal polymorphic SNPs for 1,249 individuals after rigorous quality control metrics were applied (see Supplementary Methods and Supplementary Tables 1 and 2). We compared the imputation performance of the DCEG Reference Set to that of the International HapMap and 1000 Genomes reference sets, which are available from the IMPUTE2 website (see URLs). We assessed imputation accuracy by taking directly genotyped SNP data from the DCEG Reference Set and masking subsets to simulate data from two low-cost commercial genotyping arrays commonly used in GWAS studies (Illumina Human Hap660 and Human OmniExpress). Probabilistic genotypes were imputed using both IMPUTE2 (ref. 8) and BEAGLE9 software and compared with the masked genotyped SNPs. Accuracy was measured using the squared Pearson correlation coefficient (R2) under an allelic dosage model (see Supplementary Methods). Using the new reference set, we observed higher imputation accuracy than that achieved with the combination of 1000 Genomes and HapMap data across a spectrum of minor allele frequencies (MAFs) (Fig. 1). Accuracy in individuals of European ancestry imputed from Hap660 or OmniExpress arrays, measured by the proportion of variants imputed with R2 > 0.8, improved by 34%, 23% and 12% for variants with MAFs of 3%, 5% and 10%, respectively. We estimated the difference in power to detect associations in GWAS design between an imputed data set and one composed of directly genotyped SNPs with the DCEG Reference Set by adapting a model developed by Park et al.10. When using Hap660 data for imputation, we observed detection rates of 92.9% when imputing with the DCEG Reference Set and 84.7% with the 1000 Genomes and HapMap reference sets relative to the detection rate attained with directly genotyped SNPs; for OmniExpress data, we observed detection rates of 93.9% and 86.2% for these reference sets, respectively. Because imputation accuracy depends on the similarity of haplotypes between reference and study populations, we examined an extreme scenario in which we used a reference population from Finland (Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study, ATBC) to impute genotypes using OmniExpress data from a US population of European ancestry (PLCO) (Supplementary Fig. 1). For common SNPs, there was minimal loss of imputation accuracy when using the reference population from Finland relative to the US-based Cancer Prevention Study II (CPSII) or a combined population of HapMap individuals from Utah of Northern and Western European ancestry (CEU) and from northern Italy (Toscans in Italy, TSI). This result suggests that, for common variants, a reference set of sufficient size can adequately predict common SNPs when there is a discrepancy in population ancestry, provided that comparable haplotypes are sufficiently represented. This observation should enable investigators to proceed more confidently with imputation without additional genotyping in related but not identical populations. Improved imputation of common and uncommon snps with a new reference set