Web-based phenotyping yields replication of genetic associations with sensitivity to warfarin. J. L. Mountain1, A. K. Kiefer1, M. Mullins1, T. K. Acquaye1, C. B. Marsh2, J. A. Johnson3, H. L. McLeod4, J. Y. Tung1, N. Eriksson1, K. E. Barnholt1 1) 23andMe, Inc, Mountain View, CA; 2) College of Medicine, The Ohio State University, Columbus, OH; 3) Department of Pharmacotherapy and Translational Research, University of Florida, Gainesville, FL; 4) Eshelman School of Pharmacy, University of North Carolina, Chapel Hill, NC.

   Progress in pharmacogenomics research has been hampered by the cost and time required to assemble, assess, and genotype sufficiently large cohorts of patients. 23andMes innovative research platform leverages involvement of a large and expanding cohort of over 150,000 genotyped individuals who have the option to respond to web-based questionnaires on an ongoing basis. The primary goal of this study is to develop, assess and implement web-based surveys for the rapid collection of drug response and toxicity data from thousands of individuals taking three classes of medication: warfarin, proton pump inhibitors (PPIs), and non-steroidal anti-inflammatory drugs (NSAIDs). In the first phase of this two-part study we assessed how well web-based questionnaires elucidate drug response data by comparing online self-reported data to responses obtained via semi-structured telephone interviews. Web-based surveys performed well in terms of obtaining medication response and side effects information. Test-retest reliability for most questions was high (>70%). Reporting for prescription medications taken currently proved more reliable than for medications taken in the past and for medications taken as needed. The second phase of this project is focused on validation of known associations between warfarin sensitivity and variants in the CYP2C9 and VKORC1 genes, between reactions to NSAIDs and variants in the CYP2C9 gene, and between reactions to PPIs and variants in the CYP2C19 gene. In our first analysis we examined warfarin sensitivity and CYP2C9 and VKORC1 genotypes in 1028 individuals who had taken or were taking warfarin. A test of association between reported warfarin sensitivity and sensitivity predicted on the basis of combinations of CYP2C9 (rs1057910, rs1799853) and VKORC1 (rs9923231) variants indicated borderline significance. A more detailed analysis revealed highly significant associations between self-reported dosage and rs1799853 and rs9923231, along with a near significant association with rs1057910. These initial results provide independent evidence that web-based surveys can yield drug response data of sufficiently high quality to enable highly scalable pharmacogenomics studies. The final phase of this project will leverage 23andMes customized genotyping chip to test for novel genes associated with drug metabolism, efficacy, and toxicity. This study is funded in part by NIH grant 1R43HG005807-01.