Matthew L. Freedman

Assessing the impact of population stratification on genetic association studies

Population stratification refers to differences in allele frequencies between cases and controls due to systematic differences in ancestry rather than association of genes with disease. It has been proposed that false positive associations due to stratification can be controlled by genotyping a few dozen unlinked genetic markers. To assess stratification empirically, we analyzed data from 11 case-control and case-cohort association studies. We did not detect statistically significant evidence for stratification but did observe that assessments based on a few dozen markers lack power to rule out moderate levels of stratification that could cause false positive associations in studies designed to detect modest genetic risk factors. After increasing the number of markers and samples in a case-cohort study (the design most immune to stratification), we found that stratification was in fact present. Our results suggest that modest amounts of stratification can exist even in well designed studies.

Multiple regions within 8q24 independently affect risk for prostate cancer

After the recent discovery that common genetic variation in 8q24 influences inherited risk of prostate cancer, we genotyped 2,973 SNPs in up to 7,518 men with and without prostate cancer from five populations. We identified seven risk variants, five of them previously undescribed, spanning 430 kb and each independently predicting risk for prostate cancer (P = 7.9 × 10−19 for the strongest association, and P < 1.5 × 10−4 for five of the variants, after controlling for each of the others). The variants define common genotypes that span a more than fivefold range of susceptibility to cancer in some populations. None of the prostate cancer risk variants aligns to a known gene or alters the coding sequence of an encoded protein.

The 8q24 cancer risk variant rs6983267 shows long-range interaction with MYC in colorectal cancer

An inherited variant on chromosome 8q24, rs6983267, is significantly associated with cancer pathogenesis. We present evidence that the region harboring this variant is a transcriptional enhancer, that the alleles of rs6983267 differentially bind transcription factor 7-like 2 (TCF7L2) and that the risk region physically interacts with the MYC proto-oncogene. These data provide strong support for a biological mechanism underlying this non-protein-coding risk variant.

Admixture mapping identifies 8q24 as a prostate cancer risk locus in African-American men

A whole-genome admixture scan in 1,597 African Americans identified a 3.8 Mb interval on chromosome 8q24 as significantly associated with susceptibility to prostate cancer [logarithm of odds (LOD) = 7.1]. The increased risk because of inheriting African ancestry is greater in men diagnosed before 72 years of age (P < 0.00032) and may contribute to the epidemiological observation that the higher risk for prostate cancer in African Americans is greatest in younger men (and attenuates with older age). The same region was recently identified through linkage analysis of prostate cancer, followed by fine-mapping. We strongly replicated this association (P < 4.2 × 10−9) but find that the previously described alleles do not explain more than a fraction of the admixture signal. Thus, admixture mapping indicates a major, still-unidentified risk gene for prostate cancer at 8q24, motivating intense work to find it.

STrengthening the REporting of Genetic Association Studies (STREGA)--an extension of the STROBE statement

Julian Little and colleagues present the STREGA recommendations, which are aimed at improving the reporting of genetic association studies.

Demonstrating stratification in a European American population

Population stratification occurs in case-control association studies when allele frequencies differ between cases and controls because of ancestry. Stratification may lead to false positive associations, although this issue remains controversial. Empirical studies have found little evidence of stratification in European-derived populations, but potentially significant levels of stratification could not be ruled out. We studied a European American panel discordant for height, a heritable trait that varies widely across Europe. Genotyping 178 SNPs and applying standard analytical methods yielded no evidence of stratification. But a SNP in the gene LCT that varies widely in frequency across Europe was strongly associated with height (P < 10−6). This apparent association was largely or completely due to stratification; rematching individuals on the basis of European ancestry greatly reduced the apparent association, and no association was observed in Polish or Scandinavian individuals. The failure of standard methods to detect this stratification indicates that new methods may be required.

Principles for the post-GWAS functional characterization of cancer risk loci

Genome wide association studies (GWAS) have identified more than 200 mostly new common low-penetrance susceptibility loci for cancers. The predicted risk associated with each locus is generally modest (with a per-allele odds ratio typically less than 2) and so, presumably, are the functional effects of individual genetic variants conferring disease susceptibility. Perhaps the greatest challenge in the ‘post-GWAS’ era is to understand the functional consequences of these loci. Biological insights can then be translated to clinical benefits, including reliable biomarkers and effective strategies for screening and disease prevention. The purpose of this article is to propose principles for the initial functional characterization of cancer risk loci, with a focus on non-coding variants, and to define ‘post-GWAS’ functional characterization. By December 2010, there were 1,212 published GWAS studies1 reporting significant (P < 5 × 10−8) associations for 210 traits (Table 1), and the Catalog of Published GWAS states that by March 2011, 812 publications reported 3,977 SNP associations1. This is likely a small fraction of the common susceptibility loci of low penetrance that will eventually be identified. Despite these successes in identifying risk loci, the causal variant and/or the molecular basis of risk etiology has been determined for only a small fraction of these associations2–4. Plausible candidate genes can be based on proximity to risk loci, but few have so far been defined in a more systematic manner (Supplementary Table 1). Table 1 The genomic context in which a variant is found can be used as preliminary functional analysis Increased investment in post-GWAS functional characterization of risk loci5 has now been advocated across diseases and for cardiovascular disease and diabetes6. For cancer biology, the complex interplay between genetics and the environment in many cancers poses a particularly exciting challenge for post-GWAS research. Here we suggest a systematic strategy for understanding how cancer-associated variants exert their effects. We mostly refer to SNPs throughout the paper, but we recognize that other types of common genetic (for example, copy number variants) or epigenetic variation may influence risk. Our understanding of the way in which a risk variant initiates disease pathogenesis progresses from statistical association between genetic variation and trait or disease variation to functionality and causality. The functional consequences of variants in protein-coding regions causing most monogenic disorders are more readily interpreted because we know the genetic code. For non-Mendelian or multifactorial traits, most of the common DNA variants have so far mapped to non-protein–coding regions2, where our understanding of functional consequences and causality is more rudimentary. Our hypothesis is that the trait-associated alleles exert their effects by influencing transcriptional output (such as transcript levels and splicing) through multiple mechanisms. We emphasize appropriate assays and models to test the functional effects of both SNPs and genes mapping to cancer predisposition loci. Although much of what is written is applicable to alleles discovered for any trait, the section on modeling gene effects will emphasize measuring cancer-related phenotypes. At some loci, multiple, independently associated risk alleles rather than single risk alleles may be functionally responsible for the occurrence of disease. Genotyping susceptibility loci (and their correlated variants) in multiple populations with different linkage disequilibrium (LD) structures may prove effective in substantially reducing the number of potentially causative variants (that is, the same causal variant may segregate in multiple populations), as shown for the FGFR2 locus in breast cancer7, but for most loci there will remain a set of potentially causative variants that cannot be separated at the statistical level from case-control genotype data. A susceptibility locus should be re-sequenced to ascertain all genetic variation, identifying candidate functional or causal variants and identifying candidate causal genes. Ideally, the identification of a causal SNP would be the next step to reveal the molecular mechanisms of risk modification. Practically, however, it is unclear what the criteria for causality should be, particularly in non-protein–coding regions. Thus, although we propose a framework set of analyses (Box 1), we acknowledge that the techniques and methods will continue to evolve with the field. Box 1 Strategies to progress from tag SNP to mechanism Target resequencing efforts using linkage disequilibrium (LD) structure. Use other populations to refine LD regions (for example African ancestry with shorter LD and more heterogeneity). Determine expression levels of nearby genes as a function of genotype at each locus (eQTL). Characterize gene regulatory regions by multiple empirical techniques bearing in mind that these are tissue and context specific. Combine regulatory regions with risk loci using coordinates from multiple reference genomes to capture all variation within the shorter regulatory regions that correlates with the tag SNP at each locus. Multiple experimental manipulations in model systems are needed to progressively implicate transcription units (genes) in mechanisms relevant to the associated loci: Knockouts of regulatory regions in animal (difficult and may be limited by functional redundancy, but new targeting methods in rat are promising) models followed by genome-wide expression analysis. Use chromatin association methods (3C, CHIA-PET) of regulatory regions to determine the identity of target genes (compare with eQTL data). Targeted gene perturbations in somatic cell models. Explore fully genome-wide eQTL and miRNA quantitative variation correlation in relevant tissues and cells. Explore epigenetic mechanisms in the context of genome-wide genetic polymorphism. Employ cell models and tissue reconstructions to evaluate mechanisms using gene perturbations and polymorphic variants. The human cancer cell xenograft has re-emerged as a minimal in vivo validation of these models. Above all, resist the temptation to equate any partial functional evidence as sufficient. Published claims of functional relevance should be fully evaluated using the steps detailed above.

8q24 prostate, breast, and colon cancer risk loci show tissue-specific long-range interaction with MYC

The 8q24 gene desert contains risk loci for multiple epithelial cancers, including colon, breast, and prostate. Recent evidence suggests these risk loci contain enhancers. In this study, data are presented showing that each risk locus bears epigenetic marks consistent with enhancer elements and forms a long-range chromatin loop with the MYC proto-oncogene located several hundred kilobases telomeric and that these interactions are tissue-specific. We therefore propose that the 8q24 risk loci operate through a common mechanism—as tissue-specific enhancers of MYC.

Modeling and E-M Estimation of Haplotype-Specific Relative Risks from Genotype Data for a Case-Control Study of Unrelated Individuals

The US National Cancer Institute has recently sponsored the formation of a Cohort Consortium (http://2002.cancer.gov/scpgenes.htm) to facilitate the pooling of data on very large numbers of people, concerning the effects of genes and environment on cancer incidence. One likely goal of these efforts will be generate a large population-based case-control series for which a number of candidate genes will be investigated using SNP haplotype as well as genotype analysis. The goal of this paper is to outline the issues involved in choosing a method of estimating haplotype-specific risk estimates for such data that is technically appropriate and yet attractive to epidemiologists who are already comfortable with odds ratios and logistic regression. Our interest is to develop and evaluate extensions of methods, based on haplotype imputation, that have been recently described (Schaid et al., Am J Hum Genet, 2002, and Zaykin et al., Hum Hered, 2002) as providing score tests of the null hypothesis of no effect of SNP haplotypes upon risk, which may be used for more complex tasks, such as providing confidence intervals, and tests of equivalence of haplotype-specific risks in two or more separate populations. In order to do so we (1) develop a cohort approach towards odds ratio analysis by expanding the E-M algorithm to provide maximum likelihood estimates of haplotype-specific odds ratios as well as genotype frequencies; (2) show how to correct the cohort approach, to give essentially unbiased estimates for population-based or nested case-control studies by incorporating the probability of selection as a case or control into the likelihood, based on a simplified model of case and control selection, and (3) finally, in an example data set (CYP17 and breast cancer, from the Multiethnic Cohort Study) we compare likelihood-based confidence interval estimates from the two methods with each other, and with the use of the single-imputation approach of Zaykin et al. applied under both null and alternative hypotheses. We conclude that so long as haplotypes are well predicted by SNP genotypes (we use the R2h criteria of Stram et al. [1]) the differences between the three methods are very small and in particular that the single imputation method may be expected to work extremely well.

Transferability of tag SNPs in genetic association studies in multiple populations

A general question for linkage disequilibrium–based association studies is how power to detect an association is compromised when tag SNPs are chosen from data in one population sample and then deployed in another sample. Specifically, it is important to know how well tags picked from the HapMap DNA samples capture the variation in other samples. To address this, we collected dense data uniformly across the four HapMap population samples and eleven other population samples. We picked tag SNPs using genotype data we collected in the HapMap samples and then evaluated the effective coverage of these tags in comparison to the entire set of common variants observed in the other samples. We simulated case-control association studies in the non-HapMap samples under a disease model of modest risk, and we observed little loss in power. These results demonstrate that the HapMap DNA samples can be used to select tags for genome-wide association studies in many samples around the world.