Statistical Optimization of Pharmacogenomics Association Studies: Key Considerations from Study Design to Analysis

Abstract

Research in human genetics and genetic epidemiology has grown significantly over the previous decade, particularly in the field of pharmacogenomics. Pharmacogenomics presents an opportunity for rapid translation of associated genetic polymorphisms into diagnostic measures or tests to guide therapy as part of a move towards personalized medicine. Expansion in genotyping technology has cleared the way for widespread use of whole-genome genotyping in the effort to identify novel biology and new genetic markers associated with pharmacokinetic and pharmacodynamic endpoints. With new technology and methodology regularly becoming available for use in genetic studies, a discussion on the application of such tools becomes necessary. In particular, quality control criteria have evolved with the use of GWAS as we have come to understand potential systematic errors which can be introduced into the data during genotyping. There have been several replicated pharmacogenomic associations, some of which have moved to the clinic to enact change in treatment decisions. These examples of translation illustrate the strength of evidence necessary to successfully and effectively translate a genetic discovery. In this review, the design of pharmacogenomic association studies is examined with the goal of optimizing the impact and utility of this research. Issues of ascertainment, genotyping, quality control, analysis and interpretation are considered.

Figure 1

Example of treatment dependent and treatment differential study designs answering a harmacogenomic question related to antiepileptic drug efficacy.

Figure 2

The plots of the top two principal component vectors resulting from principal components analysis before and after removal of outliers and with true race coded by color. A) Prior to exclusion of outliers, there are several individuals who do not cluster cleanly. B) After outlier removal, the presence of two general clusters can be visualized. Open Diamond = African American; Open Circle = European American; Boxed Plus Sign = Asian; Boxed X = Hispanic; Plus Sign = Other; Filled Triangle = Unknown.

Figure 2

The plots of the top two principal component vectors resulting from principal components analysis before and after removal of outliers and with true race coded by color. A) Prior to exclusion of outliers, there are several individuals who do not cluster cleanly. B) After outlier removal, the presence of two general clusters can be visualized. Open Diamond = African American; Open Circle = European American; Boxed Plus Sign = Asian; Boxed X = Hispanic; Plus Sign = Other; Filled Triangle = Unknown.

Figure 3

Q-Q plots showing the negative log-transformed p-values for all SNPs in an association study. A) Before quality control, many association results have low p-values. B) Subsequent to quality control, it can be seen that the presence of a departure from expected p-values was likely the result of systematic bias from low quality data. The line demonstrates what would be expected if the null hypothesis of no association holds.

Figure 3

Q-Q plots showing the negative log-transformed p-values for all SNPs in an association study. A) Before quality control, many association results have low p-values. B) Subsequent to quality control, it can be seen that the presence of a departure from expected p-values was likely the result of systematic bias from low quality data. The line demonstrates what would be expected if the null hypothesis of no association holds.

Figure 4

A flow chart of potential analysis designs to pursue based on the form of the study question.

Figure 5

Regression equations used in genetic epidemiology. A) The identity link function for linear regression predicts the mean trait value, Y. B) The logit link for logistic regression predicts the odds, [p/(1-p)], of the outcome. C) Poisson regression uses the log function to predict the rate, λ, of events. D) Cox proportional hazards regression also uses the log function but predicts the hazard, h(t), at time t. Each regression function assumes an intercept B₀ - the value of the mean, odds, rate or hazard when all independent variables are coded to 0 - and fits a coefficient B_n to describe the effect of each independent variable x_n.

Figure 6

Coding conventions for genetic markers with two alleles when performing regression.

Figure 7

Analysis methods used to explore gene-gene interactions.