(693b) Quantifying Synergy Effects of Multiple Drugs On Gene Expression in High-Throughput Experiments

Combination drug treatment, where two or more drugs are administered simultaneously, has been shown clinically to present therapeutic advantages in treating a variety of diseases, including cancer, neurologic disorders and infectious diseases [1]. The treatment is especially advantageous when therapeutic synergy is achieved, such that the effect of individual administration of drugs is measurably magnified above that of additivity. Because the drugs can then be co-administered at lower doses, this can result in equivalent efficacy with diminished side effects and reduced probability of drug resistance.

There is a clear need for elucidation of the molecular mechanisms underlying synergy in the development of predictive tools that would serve in treatment selection and optimization. The need for the use of mathematical, computational and engineering approaches in the quantification and analysis of drug synergy is highlighted by the very large amounts of relevant data that are produced when investigating the biological consequences of drug administration at the cellular level using high-throughput gene expression array or RNA-seq studies. Rudimentary approaches are often used in current analyses e.g., simple Venn diagrams of gene expression results from the use of each drug alone and in combination, where synergy is inferred if a gene is differentially expressed in combination drug treated samples but not when drugs are administered alone, as compared to controls.

We have addressed these problems with a systematic and rigorous analysis strategy that quantifies the impact of drug synergy, in a manner that makes it possible to elucidate the mechanistic underpinnings of the superior efficacy and reduced toxicity of synergistic drug combination therapies. Our analysis offers three notable advantages: First, it leads to quantification and classification of the degree of synergy. Second, it provides insights on antagonism of drug combinations on the expression of a particular gene. Third, the method does not result in a significant number of false positives and negatives, as it does not involve arbitrary filter thresholds (e.g. fold change, p-value cut-offs).

We quantify synergy by extending a method that is used to estimate additive, synergistic and antagonistic responses in combination drug treatment measured in enzymatic, cellular and whole animal systems to high-throughput transcriptome data [2]. Based on this analysis, we group the genes related to synergy according to functional information about them, and map them into known molecular interaction networks, signaling pathways and functional groups. The systems information introduced from the informatics databases, lead to inferences on novel genes and mechanistic roles in molecular processes affected by synergy. For example, high levels of synergy in functionally related genes indicates possible modification of a molecular mechanism in which they are involved - such as a metabolic or signaling pathway.

We have already applied an initial version of this approach to understand the synergistic effects of two anticancer drugs [3], analyzing the gene expression pattern in whole-genome arrays, affected signaling pathways and cellular pathways, in an ovarian cancer cell line (1A9) [4]. In addition to the mechanistic insights that are of special importance in cancer, the results also suggest that the majority of genes are monotonically induced or repressed with respect to the single, or combination drug-dose titrations in the therapeutic range. Additionally, the approach specifies the mode in which they are best verified experimentally.

We now investigate the applicability of the method to quantify and analyze synergy of drug combinations on the cellular transcriptome when targeting cancer and other chronic diseases. First, we will introduce the systematic framework, including the mathematical formulation of synergy and the integration with database information. Second, we will illustrate the results from our anticancer study. Third, we will show that the method is applicable to other illnesses where drug synergy effects on the transcriptome are studied in the treatment of disease pathology.

[1] Lehar J, Krueger AS, Avery W, Heilbut AM, Johansen LM, Price ER, Rickles RJ, Short GF, Staunton JE, Jin X, Lee MS, Zimmermann, and Borisky AA. Nature Biotechnology, 27: 659-66, 2009.

[2] Chou, TC and Talalay P. Adv. Enzyme. Regul. 22: 27-55, 1984.

[3] Gumus, ZH, Siso-Nadal F, Gjrezi A, McDonagh P, Khalil I, Giannakakou P, Weinstein H. IEEE Bioinformatics & Bioengineering, 2010.

[4] Marcus AI, O'Brate AM, Buey RM, Zhou J, Thomas S, Khuri FR, Andreu JM, Diaz F, Giannakakou, P. Cancer Research. 66:8838-46, 2006.