Design Of Drug Combinations Research Articles

Methods to Profile the Macromolecular Targets of Small Compounds.

Small compounds constitute most of the available medicines. However, their stereophysical and stereochemical properties are relatively simple, which typically results in promiscuity in their interactions with human proteins. Such promiscuity has caused problems in our past efforts to discover and develop new drugs, but at the same time, it also brought us new opportunities to exploit the off-target interactions between small compounds and human proteins for novel or improved therapeutics, e.g. in applications of polypharmacology, drug repositioning, and rational design of drug combinations. In this direction, identifying the full profile of macromolecules that a small compound may interact with is of fundamental importance to harnessing the positive side of small compound promiscuity. This review summarizes available experimental and computational approaches that identify macromolecular targets for small compounds. The principle, application, performance, limitation and availability of these approaches are discussed.

Modeling of signaling crosstalk-mediated drug resistance and its implications on drug combination.

The efficacy of pharmacological perturbation to the signaling transduction network depends on the network topology. However, whether and how signaling dynamics mediated by crosstalk contributes to the drug resistance are not fully understood and remain to be systematically explored. In this study, motivated by a realistic signaling network linked by crosstalk between EGF/EGFR/Ras/MEK/ERK pathway and HGF/HGFR/PI3K/AKT pathway, we develop kinetic models for several small networks with typical crosstalk modules to investigate the role of the architecture of crosstalk in inducing drug resistance. Our results demonstrate that crosstalk inhibition diminishes the response of signaling output to the external stimuli. Moreover, we show that signaling crosstalk affects the relative sensitivity of drugs, and some types of crosstalk modules that could yield resistance to the targeted drugs were identified. Furthermore, we quantitatively evaluate the relative efficacy and synergism of drug combinations. For the modules that are resistant to the targeted drug, we identify drug targets that can not only increase the relative drug efficacy but also act synergistically. In addition, we analyze the role of the strength of crosstalk in switching a module between drug-sensitive and drug-resistant. Our study provides mechanistic insights into the signaling crosstalk-mediated mechanisms of drug resistance and provides implications for the design of synergistic drug combinations to reduce drug resistance.

Tumor deconstruction as a tool for advanced drug screening and repositioning

A major focus of contemporary drug screening strategies is the identification of novel anticancer compounds, which often results in underutilization of resources. Current drug evaluation involves in vivo tumor (xenograft) regression as proof⿿of⿿principle for cytotoxicity (POC). However, this end-point lacks any assessment of drug resistance of the residual tumor and its capability to establish refractory and/or recurrent disease, which would represent more appropriate indicators of therapeutic failure. We have recently developed a flow cytometry-based approach for the analyses of intra-tumor cellular heterogeneity across stem cell hierarchies, genetic instability and differential cell cycling fractions, which can potentially be predictive of refractory disease and tumor relapse. Iterating this approach after initial POC screening in the drug discovery pipeline would have a great impact in terms of precision of drug evaluation, design of optimal drug combinations and/or drug repositioning. In this perspective, we highlight how through embracing of a comprehensive, informative and analytical assessment of the cellular content of residual tumors, the fidelity and statistical robustness of preclinical drug discovery can be greatly improved.

Abstract 788: Transcriptional landscape of drug response guides the design of potent and synergistic drug combinations

Abstract Transcriptional profiling of drug-treated cells yields high dimensional response signatures that allow drugs to be compared with each other. For example, the Connectivity Map collects signatures that are aggregated across multiple cell types. However, most therapeutic drugs are effective only against a subset of disease genotypes, particularly in the case of anti-cancer drugs. Here we ask how transcriptional signatures vary across cell lines and dose and correlate these signatures to the phenotypic response (growth inhibition). Using these cell line specific signatures, we inferred which signaling pathways are perturbed by specific kinase inhibitors and identified synergistic drug combinations. We treated 6 breast cancer cell lines with more than 100 targeted inhibitors at 6 doses and measured their transcriptional response at 2 time points. We focused on inhibitors targeting key the PI3K and MAPK signaling pathways, as well as receptor tyrosine kinases (RTKs) and cyclin-dependent kinases (CDKs); many of them are currently studied in clinical trials. We identified that ∼40% of the perturbations induce a significant difference in their gene expression profile. Clustering revealed the signatures are time point specific. Some clusters contain perturbations from multiple cell lines, like CDK inhibitors that down regulate genes related to the cell cycle in all six lines. In contrast, clusters comprising inhibitors of the PI3K and MAPK pathways are specific to each cell line and pathway. The perturbations induced by RTK and non-RTK inhibitors cluster with either the PI3K or the MAPK inhibitors depending on the cell line. Thus, the transcriptional response allows us to identify differences in pathway usage between cell lines, in particular to which pathway RTKs signal predominantly. We found that the significance of the transcriptional signature is not necessarily related to growth inhibition. In particular, some inhibitors have little effect on growth, yet induce a significant transcriptional signature. The most striking case is the inhibition of MEK and EGFR in BT20 that induces strong transcriptional and biochemical responses but inhibits growth by only ∼20%. Based on the transcriptional signature we inferred and validated experimentally that FoxO, which is generally regulated by the PI3K pathway, is partially activated by MEK or EGFR inhibition. This suggests that EGFR and PI3K inhibitors act synergistically in BT20, which we validated experimentally both at the level of FoxO activation and growth inhibition. We validated the most promising drug pair by treating xenografts. We have shown how we can use measurements of expression signatures and cellular phenotypes following single drug perturbations to identify drug combinations that are synergistic in individual cell lines. This approach is a step toward the rational design of co-drugging strategies with differential effect and larger therapeutic windows. Citation Format: Marc Hafner, Mario Niepel, Qiaonan Duan, Evan Paull, Josh Stuart, Aravind Subramanian, Avi Ma’ayan, Peter K. Sorger. Transcriptional landscape of drug response guides the design of potent and synergistic drug combinations. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 788.

Epidermal Growth Factor Receptor (EGFR) and its Cross-Talks with Topoisomerases: Challenges and Opportunities for Multi-Target Anticancer Drugs.

The interactions of Epidermal Growth Factor Receptor (EGFR) and topoisomerases have been seen in various cancer including brain, breast, ovarian, colorectal, gastric, etc. The studies in adenocarcinoma patients, chromogenic in situ hybridization, western blotting, receptor binding assay and electromobility shift assays, etc. threw light on the biophysical and biochemical features of EGFR and Topoisomerase cross-talks. It has been revealed that both the isomers of topoisomerase (Topo I and Topo II) interact via different mechanisms with EGFR. Topo II and HER2 share the same location i.e. 17q12-21 regions which could be a possible cause of predominant interactions seen between them. Topo I and EGFR interactions are mechanically related to the nucleolar translocation of heparenase by EGF and c-Jun. We compiled literature findings including the mechanistic interventions, signaling pathways, patents, in vitro and in vivo data of tested inhibitors and combinations in clinical trials, which provide convincing confirmations for the interactions of EGFR and topoisomerases. These interactions may be used for deriving a consistent route of mechanism, design and development of standard drug combinations and dual or multi inhibitors.

Abstract 35: Evolution of resistance to glucocorticoid receptor agonists

Abstract Chemotherapy resistance is a major obstacle to curing cancer patients. Over 90% of metastatic tumors are resistant to chemotherapy, leading to treatment failure. Although glucocorticoid receptor agonists (GRAs) are a very commonly used class of component drugs in front-line combination chemotherapies for hematopoietic cancer, our understanding of its mechanism(s) of action and resistance are very limited. Here using a preclinical murine model of Burkitt's lymphoma, we investigated the mechanism of resistance to the glucocorticoid receptor agonist dexamethasone. Dose responses of dexamethasone on the parental Eμ-myc; Arf-null cell line revealed a consistently small population of persistent survivors, suggesting a pre-existing subpopulation with innate dexamethasone resistance. To further address the evolution and selection of these survivors in the presence of GRAs, we derived resistant cell lines in vitro by treating the parental cell line with dose escalating concentrations of dexamethasone. These resistant cell lines were expectedly cross-resistant to other GRAs. However, loss of expression of the glucocorticoid receptor (GR) was not observed – ruling out one possible resistance mechanism. To explore alternative or downstream pathways that may be implicated, we subsequently acquired dose responses against a broad set of targeted kinase inhibitors, and observed modest sensitivity to the HSP90 inhibitor 17-AAG. This suggests that the resistance mechanism could be related to HSP90's role of holding ligand-free glucocorticoid receptor (GR) in the cytoplasm. We are now re-evolving the parental populations, and deriving clonal populations to confirm this phenotype. A better understanding of how hematopoietic cancer cells evolve resistance to GCAs will lead to improved design of drug combinations for prolonged survival. Citation Format: Simona Dalin, Boyang Zhao, Michael T. Hemann. Evolution of resistance to glucocorticoid receptor agonists. [abstract]. In: Proceedings of the AACR Precision Medicine Series: Integrating Clinical Genomics and Cancer Therapy; Jun 13-16, 2015; Salt Lake City, UT. Philadelphia (PA): AACR; Clin Cancer Res 2016;22(1_Suppl):Abstract nr 35.

Abstract B20: Transcriptional landscape of drug response guides the design of specific and potent drug combinations

Abstract Understanding responses to targeted agents is a key step toward the design of new therapeutic strategies that improve clinical cancer care. Here, we profiled the effects of a collection of kinase inhibitors using the L1000 transcriptional assay and combined these results with both phenotypic and biochemical response measurements to gain a more complete understanding of drug response. Using algorithms that reconstruct which signaling pathways are perturbed by specific kinase inhibitors, we identified potentially synergistic drug combinations and validated them experimentally. We treated six breast cancer cell lines with more than 100 targeted inhibitors at six doses and measured their transcriptional response at two time points. We focused on inhibitors targeting key the PI3K and MAPK signaling pathways, as well as receptor tyrosine kinases (RTKs) and cyclin-dependent kinases (CDKs); many of them are currently studied in clinical trials. We identified that 37% of the perturbations induce a significant difference in their gene expression profile based on the characteristic direction of the response. Clustering of signatures revealed they are time point specific: 3 hour signatures differ from the 24 hour ones. Some clusters contain perturbations from multiple cell lines, like CDK inhibitors that down regulate genes related to the cell cycle in all six lines. In contrast, clusters comprising inhibitors of the PI3K/AKT and MAPK pathways are specific to each cell line and pathway. The perturbations induced by RTK and non-RTK inhibitors cluster with either the PI3K or the MAPK inhibitors depending on the cell line. Thus, the transcriptional response allow us to identify differences in pathway usage between cell lines, in particular which RTK signals predominantly to the PI3K or the MAPK pathway. When we related transcriptional response to the growth inhibition after three days, we found that the strength of the transcriptional signature is not necessarily related to growth inhibition. In particular, we identified cases where inhibitors have little effect on growth, yet induce a significant transcriptional signature. The most striking case is the inhibition of MEK and EGFR in BT-20 that induces strong transcriptional and biochemical responses but only 20-30% of growth inhibition. Based on the transcriptional signature we inferred and validated experimentally that FoxO, which is generally regulated by the PI3K pathway, is partially activated following MEK or EGFR inhibition. This suggests that EGFR inhibitors and PI3K inhibitors act synergistically in BT-20, which we validated experimentally both at the level of FoxO activation and growth inhibition. We are currently verifying the most promising drug pair in xenografts. We have shown how we can use measurements of expression signatures and cellular phenotypes following single drug perturbations to identify drug combinations that are potent and specific to individual cell lines. This approach is a step toward the rational design of co-drugging strategies with differential effect and larger therapeutic windows. Citation Format: Marc Hafner, Mario Niepel, Qiaonan Duan, Evan Paull, Josh Stuart, Aravind Subramanian, Avi Ma'ayan, Peter Sorger. Transcriptional landscape of drug response guides the design of specific and potent drug combinations. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2015 Nov 5-9; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2015;14(12 Suppl 2):Abstract nr B20.

Abstract C70: Differences in signalling and phenotypic effects in melanoma between receptor tyrosine kinases that confer resistance to BRAF inhibitors

Abstract The majority of melanoma patients treated with a BRAF inhibitor (BRAFi) eventually acquire drug resistance. A major mechanism for acquired resistance is the activation of specific receptor tyrosine kinases (RTKs). We have recently shown that inherent and early adaptive resistance is associated with activation of the JNK-JUN pathway, decreased levels of the SPROUTY negative regulators and a mesenchymal-like phenotype. RTKs have been reported to be regulated by SPROUTY proteins and are able to mediate cell migration and invasion. In this study, we compared the expression and activation of specific RTKs previously linked to BRAFi resistance, within the same cellular background. EGFR, FGFR1, MET and IGF1R were constitutively expressed in the BRAFi-sensitive melanoma cell line A375. The impact of ligand activation of each of these receptors in the presence of BRAFi on proliferation, survival, signalling, morphology, expression of phenotype markers and cell migration was assessed. We also over-expressed SPROUTY2 using an inducible system, to study its role in regulating RTK-mediated BRAFi resistance. In A375 cells, overexpression and activation of either EGFR, FGFR1 or MET were able to reverse inhibition of cell proliferation induced by BRAFi, whereas IGF1R expression and activation was able to reduce cell death induced by high doses of BRAFi, but was unable to reverse inhibition of cell proliferation. The RTKs differed in their ability to activate signalling following treatment with BRAFi. In A375 cells, EGFR, FGFR1 and MET activated by their respective ligands, predominantly induced ERK signalling, whereas ligand activated IGF1R predominantly induced AKT signalling. EGFR activation conferred resistance to BRAFi by overcoming RAF-MEK-ERK pathway inhibition. Activation of EGFR was also able to prevent the early drug adaptive phenotype and reverse the BRAFi-induced changes in the expression of c-JUN, SNAI2, MITF and ZEB2, In contrast activation of IGF1R did not reverse the early drug adaptive phenotype induced by BRAFi. Combination treatment with either an IGF1R or AKT inhibitor was sufficient to overcome IGF1R-mediated survival. Enforced expression of SPROUTY2 reversed the resistance mediated by EGFR or FGFR1 activation. SPROUTY2 expression also conferred sensitivity to BRAFi in LOX-IMV1 cells that were inherently resistant to BRAFi and expressed activated EGFR and FGFR1. Together, our data suggests that RTKs vary in the signalling pathways that are induced upon ligand-activation, particularly in the context of BRAFi resistance. Our findings have clinical relevance for the rational design of drug combinations to target specific RTK signalling and overcome RTK-mediated resistance. Citation Format: Frederic Zhentao Li, Rachel Ramsdale, Amardeep Dhillon, Grant McArthur, Petranel Ferrao. Differences in signalling and phenotypic effects in melanoma between receptor tyrosine kinases that confer resistance to BRAF inhibitors. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2015 Nov 5-9; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2015;14(12 Suppl 2):Abstract nr C70.

Abstract PR16: Transcriptional landscape of drug response guides the design of specific and potent drug combinations

Abstract Characterizing the molecular effects of targeted therapies is an important step towards understanding and predicting drug efficacy in cancer. In this work, we used the L1000 assay developed at the Broad Institute to measure the transcriptional response of six breast cancer cell lines to more than 100 different targeted drugs, many of whom are in clinical trials. We focused on inhibitors targeting the most important signaling kinases such as PI3K, AKT or MAPK, as well as receptor tyrosine kinases (RTKs) and cyclin-dependent kinases (CDKs). With two time points and six doses, the dataset contains more than 8000 unique perturbations. We clustered the perturbations that elicited a significant response (37% of measured perturbations when using p<0.05) according to their gene expression profile and obtained 23 clusters. The perturbation significance was correlated with inhibitor dose, but no cluster was biased toward a particular concentration. Clusters were generally time point specific: the transcriptional responses at 3 hours differed significantly from the 24 hour ones. Some clusters contained perturbations from multiple cell lines whereas others were cell line specific. In particular, responses to CDK inhibitors were similar across most cell lines and showed a down-regulation of genes related to the cell cycle. On the other hand, cell lines responded differently to PI3K/AKT and MAPK inhibitors as illustrated by clusters specific to each cell line and pathway. Interestingly, the perturbations induced by RTK (e.g. EGFR, MET, ALK) and non-RTK (e.g. SRC, ABL, BTK) inhibitors, clustered with either the PI3K/AKT or the MAPK inhibitors depending on the cell line. Thus the transcriptional response allowed us to identify differences in pathway connectivity between cell lines, in particular which RTK connects to the PI3K/AKT pathway or the MAPK one. In parallel to the transcriptional response, we measured the growth inhibition after three days of treatment. We found diverse phenotypic responses that are not necessarily related to the strength of the transcriptional signature. In particular, we identified cases where inhibitors had little effect on growth, yet induced a significant transcriptional response. These cases suggest possible adaptation mechanisms to the inhibitors that may lead to drug resistance. We followed these cases experimentally and identified cell line-specific resistance mechanisms. Finally, we used the directionality of the transcriptional response in the mRNA space to guide co-drugging strategies. By testing if targeting parallel pathways is more potent than inhibiting twice the same pathway, we derived a systematic approach to design synergetic combinations. Because of the differential responses across cell lines, we were able to find combinations that are potent only in a specific cell line. This approach is a step toward the design of co-drugging strategies with differential effect and large therapeutic windows. In conclusion, our measurements of expression signatures and cellular phenotype across thousands of perturbations and six breast cancer cell lines allowed differential analyses that complement the C-MAP strategy based on consensus signatures. With our approach, we found connections between drugs based on transcriptional responses and paved the way for systematic design and analyses of drug synergies. Citation Format: Marc Hafner, Mario Niepel, Qiaonan Duan, Xiaodong Lu, Aravind Subramanian, Avi Ma'ayan, Peter K. Sorger. Transcriptional landscape of drug response guides the design of specific and potent drug combinations. [abstract]. In: Proceedings of the AACR Special Conference on Computational and Systems Biology of Cancer; Feb 8-11 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 2):Abstract nr PR16.

Mitochondrial Profiling of Acute Myeloid Leukemia in the Assessment of Response to Apoptosis Modulating Drugs.

BH3 profiling measures the propensity of transformed cells to undergo intrinsic apoptosis and is determined by exposing cells to BH3-mimicking peptides. We hypothesized that basal levels of prosurvival BCL-2 family proteins may modulate the predictive power of BH3 profiling and termed it mitochondrial profiling. We investigated the correlation between cell sensitivity to apoptogenic agents and mitochondrial profiling, using a panel of acute myeloid leukemias induced to undergo apoptosis by exposure to cytarabine, the BH3 mimetic ABT-199, the MDM2 inhibitor Nutlin-3a, or the CRM1 inhibitor KPT-330. We found that the apoptogenic efficacies of ABT-199 and cytarabine correlated well with BH3 profiling reflecting BCL2, but not BCL-XL or MCL-1 dependence. Baseline BCL-2 protein expression analysis increased the ability of BH3 profiling to predict resistance mediated by MCL-1. By utilizing engineered cells with overexpression or knockdown of BCL-2 family proteins, Ara-C was found to be independent, while ABT-199 was dependent on BCL-XL. BCL-2 and BCL-XL overexpression mediated resistance to KPT-330 which was not reflected in the BH3 profiling assay, or in baseline BCL-2 protein levels. In conclusion, mitochondrial profiling, the combination of BH3 profiling and prosurvival BCL-2 family protein analysis, represents an improved approach to predict efficacy of diverse agents in AML and may have utility in the design of more effective drug combinations.

Systematic synergy modeling: understanding drug synergy from a systems biology perspective.

Owing to drug synergy effects, drug combinations have become a new trend in combating complex diseases like cancer, HIV and cardiovascular diseases. However, conventional synergy quantification methods often depend on experimental dose–response data which are quite resource-demanding. In addition, these methods are unable to interpret the explicit synergy mechanism. In this review, we give representative examples of how systems biology modeling offers strategies toward better understanding of drug synergy, including the protein-protein interaction (PPI) network-based methods, pathway dynamic simulations, synergy network motif recognitions, integrative drug feature calculations, and “omic”-supported analyses. Although partially successful in drug synergy exploration and interpretation, more efforts should be put on a holistic understanding of drug-disease interactions, considering integrative pharmacology and toxicology factors. With a comprehensive and deep insight into the mechanism of drug synergy, systems biology opens a novel avenue for rational design of effective drug combinations.

Anthelmintics: The best way to predict the future is to create it

‘The best way to predict the future is to create it.’ When we look at drugs that are used to control parasites, we see that new knowledge has been created (discovered) about their modes of action. This knowledge will allow us to predict combinations of drugs which can be used together rationally to increase the spectrum of action and to slow the development of anthelmintic resistance. In this paper we comment on some recent observations of ours on the modes of action of emodepside, diethylcarbamazine and tribendimidine. Emodepside increases the activation of a SLO-1 K+ current inhibiting movement, and diethylcarbamazine has a synergistic effect on the effect of emodepside on the SLO-1 K+ current, increasing the size of the response. The combination may be considered for further testing for therapeutic use. Tribendimidine is a selective cholinergic nematode B-subtype nAChR agonist, producing muscle depolarization and contraction. It has different subtype selectivity to levamisole and may be effective in the presence of some types of levamisole resistance. The new information about the modes of action may aid the design of rational drug combinations designed to slow the development of resistance or increase the spectrum of action.

Perspectives in the development of hybrid bifunctional antitumour agents

In spite of the development of a large number of novel target-specific antitumour agents, the single-agent therapy is in general not able to provide an effective durable control of the malignant process. The limited efficacy of the available agents (both conventional cytotoxic and novel target-specific) reflects not only the expression of defence mechanisms, but also the complexity of tumour cell alterations and the redundancy of survival pathways, thus resulting in tumour cell ability to survive under stress conditions. A well-established strategy to improve the efficacy of antitumour therapy is the rational design of drug combinations aimed at achieving synergistic effects and overcoming drug resistance. An alternative strategy could be the use of agents designed to inhibit simultaneously multiple cellular targets relevant to tumour growth/survival. Among these novel agents are hybrid bifunctional drugs, i.e. compounds resulting by conjugation of different drugs or containing the pharmocophores of different drugs. This strategy has been pursued using various conventional or target-specific agents (with DNA damaging agents and histone deacetylase inhibitors as the most exploited compounds). A critical overview of the most representative compounds is provided with emphasis on the HDAC inhibitor-based hybrid agents. In spite of some promising results, the actual pharmacological advantages of the hybrid agents remain to be defined. This commentary summarizes the recent advances in this field and highlights the pharmacological basis for a rational design of hybrid bifunctional agents.

Experimental design and statistical analysis for three-drug combination studies.

Drug combination is a critically important therapeutic approach for complex diseases such as cancer and HIV due to its potential for efficacy at lower, less toxic doses and the need to move new therapies rapidly into clinical trials. One of the key issues is to identify which combinations are additive, synergistic, or antagonistic. While the value of multidrug combinations has been well recognized in the cancer research community, to our best knowledge, all existing experimental studies rely on fixing the dose of one drug to reduce the dimensionality, e.g. looking at pairwise two-drug combinations, a suboptimal design. Hence, there is an urgent need to develop experimental design and analysis methods for studying multidrug combinations directly. Because the complexity of the problem increases exponentially with the number of constituent drugs, there has been little progress in the development of methods for the design and analysis of high-dimensional drug combinations. In fact, contrary to common mathematical reasoning, the case of three-drug combinations is fundamentally more difficult than two-drug combinations. Apparently, finding doses of the combination, number of combinations, and replicates needed to detect departures from additivity depends on dose-response shapes of individual constituent drugs. Thus, different classes of drugs of different dose-response shapes need to be treated as a separate case. Our application and case studies develop dose finding and sample size method for detecting departures from additivity with several common (linear and log-linear) classes of single dose-response curves. Furthermore, utilizing the geometric features of the interaction index, we propose a nonparametric model to estimate the interaction index surface by B-spine approximation and derive its asymptotic properties. Utilizing the method, we designed and analyzed a combination study of three anticancer drugs, PD184, HA14-1, and CEP3891 inhibiting myeloma H929 cell line. To our best knowledge, this is the first ever three drug combinations study performed based on the original 4D dose-response surface formed by dose ranges of three drugs.

Searching for Drug Synergy in Complex Dose–Response Landscapes Using an Interaction Potency Model

Rational design of multi-targeted drug combinations is a promising strategy to tackle the drug resistance problem for many complex disorders. A drug combination is usually classified as synergistic or antagonistic, depending on the deviation of the observed combination response from the expected effect calculated based on a reference model of non-interaction. The existing reference models were proposed originally for low-throughput drug combination experiments, which make the model assumptions often incompatible with the complex drug interaction patterns across various dose pairs that are typically observed in large-scale dose–response matrix experiments. To address these limitations, we proposed a novel reference model, named zero interaction potency (ZIP), which captures the drug interaction relationships by comparing the change in the potency of the dose–response curves between individual drugs and their combinations. We utilized a delta score to quantify the deviation from the expectation of zero interaction, and proved that a delta score value of zero implies both probabilistic independence and dose additivity. Using data from a large-scale anticancer drug combination experiment, we demonstrated empirically how the ZIP scoring approach captures the experimentally confirmed drug synergy while keeping the false positive rate at a low level. Further, rather than relying on a single parameter to assess drug interaction, we proposed the use of an interaction landscape over the full dose–response matrix to identify and quantify synergistic and antagonistic dose regions. The interaction landscape offers an increased power to differentiate between various classes of drug combinations, and may therefore provide an improved means for understanding their mechanisms of action toward clinical translation.

Response to comments on 'competing designs for drug combination in phase I dose-finding clinical trials' by G. Yin, R. Lin and N. Wages.

As pointed out by Yin and Lin, and Wages, designs for drug-combination trials need to be rigorously evaluated. In our paper, we performed an extensive sensitivity analysis (with some of the results given in the form of additional material), and for each design, we tried to find the best parameter settings for model calibration associated with a high percentage of correct selection (PCS) on average across all scenarios.

Comments on 'competing designs for drug combination in phase I dose-finding clinical trials' by M-K. Riviere, F. Dubois, S. Zohar.

In a recent paper, Riviere, Dubois, and Zohar [1] conducted a comparative study of dose-finding methods for two-agent combination phase I oncology trials. One of the methods presented by the authors in their comparison was the partial order continual reassessment method (POCRM; [2]). Unfortunately, some of the design choices made by the authors in their implementation of POCRM hindered its performance in some scenarios. The purpose of this letter is to show how Riviere et al. [1] could have better implemented POCRM in conducting their simulation study. Specifically, the following claims are addressed and further clarification provided to ensure that accurate implementation of POCRM is followed. “As most real clinical trials begin at the lowest dose, we think that this method (POCRM) should be modified to allow initiating the trial at the lowest combination while adding ‘nonskipping’ rules or a start-up phase.” POCRM has already been modified to include a start-up phase that begins at the lowest dose combination, and work that was published by Wages, Conaway and O’Quigley [3] in 2011 evaluated the operating characteristics of this modification. POCRM was the only model-based method of those compared by Riviere et al. [1] in which the authors chose not to include a start-up phase, even though a modified version of POCRM that includes such a phase has already been published [3], and is accompanied by publically available software [4]. “Another important modification that we have added to the POCRM is the recommendation to start at the lowest dose. We have based our decision on common practice in Phase I for a single agent or a combination of agents.” As stated above, POCRM has already been modified to begin at the lowest dose combination [3]. In contrast, the modification made to POCRM by Riviere et al. [1] was not only to begin at the lowest dose of each agent (i.e. combination D1,1), but to do so within a Bayesian framework using priors and skeletons that center the prior MTD at combinations D3,3, D4,3 or D5,2, depending on the dose-toxicity ordering. After beginning the trial at D1,1, in the absence of DLT in the first cohort, this modification will cause the design to gravitate toward the prior MTD (i.e. D3,3, D4,3 or D5,2), and such behavior would not be allowed in practice, as pointed out by the authors. Therefore, the modification made by the authors does not resolve any practical issue associated with the POCRM. The authors have chosen to make a modification that is not allowable in practice within an article that is meant to inform statisticians about how best to practically conduct Phase I trials of combinations. Wages and Conaway [5] published practical guidelines for implementing POCRM, yet the comparison by Riviere et al. [1] did not follow these recommendations, especially in choosing an appropriate subset of possible dose-toxicity orderings. The guidelines in [5], combined with an R package (pocrm [4]), have led to the successful implementation of POCRM in two ongoing trials since its publication in 2011. One recently completed accrual and is in follow-up (Mel 58) and the other (Mel 60) is open to accrual at two sites, the University of Virginia Cancer Center and MD Anderson Cancer Center. Descriptions of these trials can be found at https://clinicaltrials.gov/ct2/show/{type:clinical-trial,attrs:{text:NCT01585350,term_id:NCT01585350}}NCT01585350?term=mel58&rank=1 and https://clinicaltrials.gov/ct2/show/{type:clinical-trial,attrs:{text:NCT02126579,term_id:NCT02126579}}NCT02126579?term=mel60&rank=1. The use of appropriate POCRM specifications in the comparative work of Riviere et al. [1] would have resulted in a more accurate evaluation of POCRM’s design characteristics, and provided the readers of Statistics in Medicine valuable insight on how the method fares relative to existing methods. In this letter, we focus our attention on proper selection of possible dose-toxicity orderings for use in the POCRM. In the subsequent sections, we review the recommendations of Wages and Conaway [5] in choosing orderings, and discuss their impact on the operating characteristics of POCRM. This sheds a very different light on the results and conclusions of Riviere [1].

Comments on 'competing designs for drug combination in phase I dose-finding clinical trials' by M-K. Riviere, F. Dubois, and S. Zohar.

Comments on 'competing designs for drug combination in phase I dose-finding clinical trials' by M-K. Riviere, F. Dubois, and S. Zohar.

Intratumor heterogeneity alters most effective drugs in designed combinations.

The substantial spatial and temporal heterogeneity observed in patient tumors poses considerable challenges for the design of effective drug combinations with predictable outcomes. Currently, the implications of tissue heterogeneity and sampling bias during diagnosis are unclear for selection and subsequent performance of potential combination therapies. Here, we apply a multiobjective computational optimization approach integrated with empirical information on efficacy and toxicity for individual drugs with respect to a spectrum of genetic perturbations, enabling derivation of optimal drug combinations for heterogeneous tumors comprising distributions of subpopulations possessing these perturbations. Analysis across probabilistic samplings from the spectrum of various possible distributions reveals that the most beneficial (considering both efficacy and toxicity) set of drugs changes as the complexity of genetic heterogeneity increases. Importantly, a significant likelihood arises that a drug selected as the most beneficial single agent with respect to the predominant subpopulation in fact does not reside within the most broadly useful drug combinations for heterogeneous tumors. The underlying explanation appears to be that heterogeneity essentially homogenizes the benefit of drug combinations, reducing the special advantage of a particular drug on a specific subpopulation. Thus, this study underscores the importance of considering heterogeneity in choosing drug combinations and offers a principled approach toward designing the most likely beneficial set, even if the subpopulation distribution is not precisely known.

Synergistic and Antagonistic Drug Combinations Depend on Network Topology

Drug combinations may exhibit synergistic or antagonistic effects. Rational design of synergistic drug combinations remains a challenge despite active experimental and computational efforts. Because drugs manifest their action via their targets, the effects of drug combinations should depend on the interaction of their targets in a network manner. We therefore modeled the effects of drug combinations along with their targets interacting in a network, trying to elucidate the relationships between the network topology involving drug targets and drug combination effects. We used three-node enzymatic networks with various topologies and parameters to study two-drug combinations. These networks can be simplifications of more complex networks involving drug targets, or closely connected target networks themselves. We found that the effects of most of the combinations were not sensitive to parameter variation, indicating that drug combinational effects largely depend on network topology. We then identified and analyzed consistent synergistic or antagonistic drug combination motifs. Synergistic motifs encompass a diverse range of patterns, including both serial and parallel combinations, while antagonistic combinations are relatively less common and homogenous, mostly composed of a positive feedback loop and a downstream link. Overall our study indicated that designing novel synergistic drug combinations based on network topology could be promising, and the motifs we identified could be a useful catalog for rational drug combination design in enzymatic systems.