Methodological and statistical issues in pharmacogenomics

Abstract

Pharmacogenomics strives to explain the interindividual variability in response to drugs due to genetic variation. Although technological advances have provided us with relatively easy and cheap methods for genotyping, promises about personalised medicine have not yet met our high expectations. Successful results that have been achieved within the field of pharmacogenomics so far are, to name a few, HLA-B*5701 screening to avoid hypersensitivity to the antiretroviral abacavir, thiopurine S-methyltransferase (TPMT) genotyping to avoid thiopurine toxicity, and CYP2C9 and VKORC1 genotyping for better dosing of the anticoagulant warfarin. However, few pharmacogenetic examples have made it into clinical practice in the treatment of complex diseases. Unfortunately, lack of reproducibility of results from observational studies involving many genes and diseases seems to be a common pattern in pharmacogenomic studies. In this article we address some of the methodological and statistical issues within study design, gene and single nucleotide polymorphism (SNP) selection and data analysis that should be considered in future pharmacogenomic research. First, we discuss some of the issues related to the design of epidemiological studies, specific to pharmacogenomic research. Second, we describe some of the pros and cons of a candidate gene approach (including gene and SNP selection) and a genome-wide scan approach. Finally, conventional as well as several innovative approaches to the analysis of large pharmacogenomic datasets are proposed that deal with the issues of multiple testing and systems biology in different ways.

Human leukocyte antigen (HLA) and pharmacogenetics: screening for HLA-B*57:01 among human immunodeficiency virus–positive patients from southern Alberta

“…We strongly believe that there is a need for more such international collaborations to study HLAassociated pharmacogenetics and establish common pathways that can help in understanding the etiology of ADRs. When carrying out these kinds of pharmacogenetic studies, methodological and statistical considerations are important to assure the validity of outcomes and to successfully convert the knowledge into clinical practice [25]. North American HILs are well placed to conduct genetic testing for HLA-associated pharmacogenetics because most are performing HLA genotyping at the variable level of resolution (high and low).…”

Analytical strategies for discovery and replication of genetic effects in pharmacogenomic studies

“…As outlined in the introduction, SSAT has been used since the inception of PGx studies despite suboptimal performance in the context of translational research in the clinical trial setting (for example, in the case of relatively small sample sizes, diverse cohorts, and potential polygenic effects of drug responses) 4. Alternatively, RBAT approaches have the potential to improve power (by leveraging the underlying LD structure in genetic data and reducing the multiplicity burden) and to detect complex PGx effects.…”

Pharmacogenomic characterization of gemcitabine response – a framework for data integration to enable personalized medicine

“…The use of probabilistic networks in conjunction with traditional statistical models for mining relationships and associations from genotype-phenotype data is well established [15]. Probabilistic network methods for pharmacogenomics and newer methods such as the Markov Blanket concept may be helpful to better analyze these complex genotype-phenotype associations [16]. Considering the complexity of both cancer prognosis and individual drug response to chemotherapeutics, application of these association methods in conjunction with novel informatics and data integration approaches is necessary to identify clinically relevant variants for validation studies and ultimately testing in the clinic for pharmacogenomics applications.…”

“…Of the numerous network learning algorithms in PNA, we applied the Augmented Markov Blanket (AMB) algorithm because it has the ability to subset a limited number of SNPs that best predict the outcome [16]. A Markov Blanket corresponds to a set of nodes in the network that make the target independent of all the other nodes conditional on this subset of nodes.…”