Using Expression and Genotype to Predict Drug Response in Yeast Douglas M. Ruderfer1,2,5, David C. Roberts3,5, Stuart L. Schreiber2,4, Ethan O. Perlstein5*, Leonid Kruglyak5,6* 1 Center for Human Genetic Research, Massachusetts General Hospital, Boston, Massachusetts, United States of America, 2 Eli & Edythe L. Broad Institute of Harvard and MIT, Cambridge, Massachusetts, United States of America, 3 Theoretical Division and Center for Nonlinear Studies, Los Alamos National Laboratory, Los Alamos, New Mexico, United States of America, 4 Department of Chemistry and Chemical Biology, Howard Hughes Medical Institute, Cambridge, Massachusetts, United States of America, 5 Lewis-Sigler Institute for Integrative Genomics, Princeton University, Princeton, New Jersey, United States of America, 6 Department of Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey, United States of America Abstract Personalized, or genomic, medicine entails tailoring pharmacological therapies according to individual genetic variation at genomic loci encoding proteins in drug-response pathways. It has been previously shown that steady-state mRNA expression can be used to predict the drug response (i.e., sensitivity or resistance) of non-genotyped mammalian cancer cell lines to chemotherapeutic agents. In a real-world setting, clinicians would have access to both steady-state expression levels of patient tissue(s) and a patient���s genotypic profile, and yet the predictive power of transcripts versus markers is not well understood. We have previously shown that a collection of genotyped and expression-profiled yeast strains can provide a model for personalized medicine. Here we compare the predictive power of 6,229 steady-state mRNA transcript levels and 2,894 genotyped markers using a pattern recognition algorithm. We were able to predict with over 70% accuracy the drug sensitivity of 104 individual genotyped yeast strains derived from a cross between a laboratory strain and a wild isolate. We observe that, independently of drug mechanism of action, both transcripts and markers can accurately predict drug response. Marker-based prediction is usually more accurate than transcript-based prediction, likely reflecting the genetic determination of gene expression in this cross. Citation: Ruderfer DM, Roberts DC, Schreiber SL, Perlstein EO, Kruglyak L (2009) Using Expression and Genotype to Predict Drug Response in Yeast. PLoS ONE 4(9): e6907. doi:10.1371/journal.pone.0006907 Editor: Thomas Preiss, Victor Chang Cardiac Research Institute (VCCRI), Australia Received January 23, 2009 Accepted July 30, 2009 Published September 4, 2009 Copyright: �� 2009 Ruderfer et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: US National Institute of Mental Health (L.K.) Center grant P50GM071508 (L.K.) James S. McDonnell Centennial Fellow (L.K.)The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: leonid@genomics.princeton.edu (LK) eperlste@princeton.edu (EOP) Introduction Realizing the promise of personalized medicine ��� a rational approach to tailoring pharmacological therapy to individual patients ��� is an area of intense research [1,2]. One of the central experimental challenges of personalized medicine is to identify physiological correlates (i.e., biomarkers) of individual genetic variation that would serve as reliable diagnostic indicators of (desired) drug response and (undesired) side effects [3���5]. The most obvious diagnostic indicator of drug response is genetic variation itself in the form of single-nucleotide polymorphisms (SNPs), insertions or deletions, or gross chromosomal rearrange- ments, which can be catalogued by genotyping techniques. Genetic variation in genes known to modulate drug response in general, such as the ABC family of xenobiotic transporters, or the cytochrome P450 detoxification enzymes, has been successfully correlated to clinical outcomes of drug therapy [6���11]. Addition- ally, in candidate-gene approaches, polymorphisms in the molecular targets of drugs (or downstream pathway components) have also been correlated with clinical outcome in cancer and in other diseases [12���18]. A complementary diagnostic indicator of drug response is mRNA expression. This approach is less biased than candidate- gene approaches because it uses global transcriptional signatures of cells in the untreated state to predict drug response. Specifically, previous work, exemplified by Staunton et al., on the NCI-60 panel ��� a collection of 60 tumor cells lines of different tissue origins that has served as the primary cellular model of cancer genomics [19] ��� demonstrated that steady-state mRNA expression could be used to predict the sensitivity of cancer cell lines to anti-neoplastic drugs [20]. In related work, others have shown that different physiological correlates besides gene expression, such as protein abundance, may be used to predict drug response [21,22]. More recently, studies have exploited a collection of genotyped mammalian cell lines to interrogate drug response, and others have performed expression profiling of mammalian cells in response to drug treatment [14,17,23]. We sought to perform a drug-response prediction analysis in a well-controlled system. We have previously studied segregating variation in a panel of 104 genotyped segregants derived from a cross between a laboratory strain (BY) and a wild isolate (RM) of Saccharomyces cerevisiae [24��� 27]. In the present study we compare how well drug response is predicted by two types of variables���expression and genotype, test the potential improvement in drug-response prediction by combining more than one variable type, and evaluate why one variable type is more predictive than another for a given drug or PLoS ONE | www.plosone.org 1 September 2009 | Volume 4 | Issue 9 | e6907

class of drugs. Our approach affords an opportunity to assess the predictive power of transcripts versus genetic markers in a simple model system, and may serve as an initial point of departure for future analogous studies of personalized medicine in humans. Results Drug response can be predicted from transcript levels in untreated cells We sought to predict the response of each segregant to each small-molecule perturbagen (drug), or SMP, from patterns of gene expression measured in a neutral (i.e., SMP-free) medium. We classified each segregant as sensitive, resistant or partially resistant to a given SMP according to its final yield in that SMP 225 SMP responses were tested (this represents 89 SMPs, with multiple responses to some SMPs measured at different time points and concentrations). The gene expression levels of the segregants classified as sensitive or resistant were used to train a support vector machine (SVM) [28]. SVMs are very powerful at classifying multidimensional data, and therefore should give us the ability to predict both Mendelian and genetically complex SMP responses. For each SMP, we used a feature selection algorithm within each fold of the cross-validation approach (see Methods) to rank the genes according to their individual contribution to the ability to predict segregant SMP responses. We then trained support vector classifiers using 1, 10, 50, 100, 200, 500, and 1000 highest ranked gene(s). The classifiers were used to predict sensitive/resistant status of segregants not included in the training set using a cross- validation approach (Methods and Data S1). We found that the support vector classifier trained on 1000 genes had the greatest average predictive power, correctly predicting the SMP response of 69.7% of the segregants on average for the SMPs considered, although it should be noted that the differences between 50, 100, 200, 500 and 1000 highest ranked genes are negligible (Figure 1). The prediction accuracy for individual SMPs varied from near 100% to near chance. We compared this classifier to a na��ve �� mode classifier, using the same cross validation as with the SVM, which calls all segregants in the test set sensitive or resistant according to the category that occurs more frequently in the training set (this provides a better comparison of the predictive value of expression information than does 50:50 random classification). Taking the prediction accuracy from the best performing set of features for each compound, the SVM outperformed the mode classifier on average (74% vs. 64%) and equaled or outperformed it for all but one SMP considered (the single instance was well within the standard deviation of the SVM performance). Instances where the mode classifier performed well reflect unequal distributions of sensitive and resistant segregants. Performance was very robust Figure 1. Summary of marker-based and transcript-based prediction algorithms. Box plots representing the distribution of prediction accuracies for all SMPs plotted against number of features selected for prediction. (A) Results of marker-based expression. (B) Results of transcript- based prediction. doi:10.1371/journal.pone.0006907.g001 Yeast Pharmacogenomics Model PLoS ONE | www.plosone.org 2 September 2009 | Volume 4 | Issue 9 | e6907