Genome-wide Studies in Pharmacogenomics: Harnessing the Power of Extreme Phenotypes

Abstract

Baldwin et al. on paclitaxel-induced sensory peripheral neuropathy [5], and Brown et al. on temozolomideassociated cytotoxicity [6] have advanced our understanding of DNA sequence variations and drug response, and several pharmaco genomic tests are on their way to clinical practice [7]. The studies on toxic drug reactions in particular have extra power compared with typical GWAS. For example, the study by Daly et al. on flucloxacillin-induced liver injury, an extreme and rare drug response phenotype, included only 51 affected individuals and 282 controls, yet it reported the unprecedented odds risk ratio of 80.6 (p = 8.7 × 10 for carriers of HLA-B*5701) [4], while typical GWAS odds risk ratios are below 2. We propose that the extreme phenotypes approach should be prioritized for genomewide pharmacogenomic studies, including those exploring genomic biomarkers for treatment efficacy as an alternative to enrolling very large random patient cohorts. This is based on both the likelihood of an enrichment for genomic signal due to the ‘Mendelian’ nature of the pheno type, the favorable study costs and reduced ‘noise’ arising from inaccurate pheno typing. Taken together, it seems far more effective to compare genome-wide data between ‘excellent drug responders’ versus nonresponders, compared with analyzing such data from huge patient cohorts representing a large spectrum of intermediate responders. The superior power of the extreme phenotype approach is exemplified by the recent study of Edmond et al., who identified DCTN4 as a modifier of chronic infection in cystic fibrosis among 91 individuals whose exomes were sequenced [8]. As a further example, expression levels of CHL1 were recently reported as Four years ago this journal featured a theme issue on genome-wide association studies (GWAS) in pharmacogenomics. In our accompanying editorial we noted that “relatively few GWAS on drug response were so far published compared with the large numbers of disease risk GWAS” and called upon the pharmacogenomics research community to apply this powerful, hypothesis-free research approach for improving knowledge on drug safety and efficacy [1]. Our call seems to have been timely, as the past few years have witnessed an increase in pharmacogenomics-oriented genome-wide studies. Among PubMed-listed pharmacogenomics-related manuscripts, the fraction of those mentioning both ‘pharmacogenomics’ and ‘genome-wide’ as keywords have risen fourfold over the past few years, from 2.1% in 2006 to nearly 9% of total manuscripts published between 2009 and 2012 (Table 1). However, as in past years, only a tiny fraction (under 2%) of published genome-wide studies (in all areas) are concerned with pharmacogenomics (Table 1), a disappointingly low figure considering the huge societal burden of adverse drug events and ineffective therapeutics [2].