Cancer Systems Biology Research Articles

Challenges in Measuring Cost and Value in Oncology: Making It Personal

Oncology patients often find themselves facing an incurable disease with limited treatment options and increasing patient fragility. The importance of patient preferences and values increases in shared decision making especially when the cost of cancer care is continuing its steep rise. As our understanding of cancer systems biology increases, we are justifiably optimistic about therapeutic improvements but recognize that this has complicated the traditional Food and Drug Administration approval of drug indications based on organ-specific cancer for a particular drug. Dynamic and agile clinical guidelines that reflect a rapidly changing knowledge base for decision-making support are needed. The American Society of Clinical Oncology (ASCO) has been working on three initiatives to tackle these complex issues. The first initiative is ASCO’s collaboration with other international organizations to create a framework to assess drugs for the World Health Organization’s Essential Medicines List, including nongenerics. The second initiative aims to define clinically meaningful outcomes as precision medicine expands the definition of cancers, leading to increased demand for the use of targeted drugs as single agents or in combination. The third initiative is ASCO’s value framework, published in 2015, focusing on patient-physician shared decision making. The framework incorporates three parameters: 1) the meaningfulness of the clinical benefit, 2) the toxicity of the treatment, and 3) the patient’s financial out-of-pocket cost. ASCO is concerned about the rising cost of cancer care when the clinical complexity and the pace of change in oncology are accelerating, and it is committed to help improve patient outcomes and value in cancer care as well as to engage the broader health care community in a process of collaborative improvement.

Reconstruction, Topological and Gene Ontology Enrichment Analysis of Cancerous Gene Regulatory Network Modules

The availability of large set of high throughput biological data needs algorithm that automatically reconstructs gene regulatory networks from these datasets. Cancerous regulatory network modules when analyzed critically may reveal the underlying mechanism of cancer, which may help in better diagnosis. Identification of cancerous genes and their regulation is an important research area in cancer systems biology. In this paper, we introduced an algorithm to infer cancerous gene regulatory network modules from gene expression profiles. The proposed algorithm has been applied to gene expression dataset of colon cancer patients and several network modules have been identified. We performed topological analysis of inferred network modules in terms of network density, degree distribution, clustering coefficient, average path length, network heterogeneity, and centrality measures. Further, GO-based enrichment analysis of the inferred network has been performed. To validate the proposed algorithm, it has been tested on benchmark dataset taken from DREAM3 challenge project. Keywords: Gene regulatory network, biological network, systems biology, GO enrichment analysis, topological analysis, cancerous network modules.

Fatty acid synthase and the lipogenic phenotype in WSSV-infected shrimp

Shrimp aquaculture is a global industry that incurs severe economic losses from several serious shrimp diseases. One of the most damaging diseases is white spot disease (WSD), which originally swept through the cultured shrimp industry with devastating effects in 1995. However, the change of host metabolism after WSSV infection is not well understood. In this talk, we review some studies about the metabolic changes in WSSV-infected shrimp. We recently used an approach based on human cancer systems biology to detect global metabolic changes induced by white spot syndrome virus (WSSV) at various stages of viral replication in white shrimp (Litopenaeus vannamei). Most interestingly, at the genome replication stage (12 hpi), WSSV induced metabolic rerouting (invertebrate Warburg effect). Furthermore, using a quantitative GC-MS-based lipidomic approach, we determined that lipogenesis of shrimp cells occurred during a late stage of WSSV replication. Fluctuations in lipid concentrations in WSSV-infected cells may be tightly coupled to WSSV-induced metabolic reprogramming and therefore, important to WSSV replication. When fatty acid synthase (FAS; the key enzyme involved in lipogenesis) was suppressed with a specific inhibitor (C75), the virus was unable to complete its replication cycle, even though replication of the WSSV genome was not affected. In addition, the PI3K-Akt-mTOR pathway was the key pathway to induce lipid biosynthesis. We concluded that WSSV-induced lipid synthesis was important to support viral morphogenesis.

Abstract IA06: Spatial systems biology of cancer

Abstract International efforts have now defined the genomic landscapes of most major human cancer types. The genomic landscape of breast cancer is particularly well described in several recent publications. The high level view of this disease suggests a remarkable level of inter- and intra-tumor heterogeneity, the presence of a few recurrent and many rare genomic aberrations and substantial genomic and epigenomic heterogeneity. The challenge now is to understand how these aberrations and tumor extrinsic influences from the microenvironment collaborate to deregulate aspects of cell differentiation, proliferation, apoptosis, DNA repair, senescence, motility and multiscale architecture that include cancer progression and response to therapy. To this end, the OHSU Center for Spatial Systems Biomedicine has deployed a suite of measurement technologies that enable high throughput assessment of changes in these endpoints to siRNA knockdown of genes found by the TCGA and international genomics efforts to be aberrant and/or involved in regulatory networks that are associated with differentiation subtype and/or differential response to therapy. These studies have revealed both recurrent and rare aberrations that strongly influence these endpoints as well as a range of molecular architectures that influence aspects of receptor tyrosine kinase signal transduction. This presentation will summarize the results of the siRNA knockdown experiments and focus on the concept that analysis of the architectural features influenced by genomic and epigenomic abnormalities will suggest new therapeutic approaches. Illustrative examples of architectural features revealed using advanced imaging approaches that suggest new approaches to the design of RTK targeted therapies include (a) Ras/Raf multimers associated with signal transduction, (b) actin associated microfilaments involved in ERBB2 signal transduction, and (c) intratumor heterogeneity in differentiation status that influences responses to a range of kinase inhibitors. Citation Format: Joe Gray. Spatial systems biology of cancer. [abstract]. In: Proceedings of the AACR Special Conference on Advances in Breast Cancer Research; Oct 17-20, 2015; Bellevue, WA. Philadelphia (PA): AACR; Mol Cancer Res 2016;14(2_Suppl):Abstract nr IA06.

Multi-OMICs and Genome Editing Perspectives on Liver Cancer Signaling Networks

The advent of the human genome sequence and the resulting ~20,000 genes provide a crucial framework for a transition from traditional biology to an integrative “OMICs” arena (Lander et al., 2001; Venter et al., 2001; Kitano, 2002). This brings in a revolution for cancer research, which now enters a big data era. In the past decade, with the facilitation by next-generation sequencing, there have been a huge number of large-scale sequencing efforts, such as The Cancer Genome Atlas (TCGA), the HapMap, and the 1000 genomes project. As a result, a deluge of genomic information becomes available from patients stricken by a variety of cancer types. The list of cancer-associated genes is ever expanding. New discoveries are made on how frequent and highly penetrant mutations, such as those in the telomerase reverse transcriptase (TERT) and TP53, function in cancer initiation, progression, and metastasis. Most genes with relatively frequent but weakly penetrant cancer mutations still remain to be characterized. In addition, genes that harbor rare but highly penetrant cancer-associated mutations continue to emerge. Here, we review recent advances related to cancer genomics, proteomics, and systems biology and suggest new perspectives in targeted therapy and precision medicine.

Combinatorial PX-866 and Raloxifene Decrease Rb Phosphorylation, Cyclin E2 Transcription, and Proliferation of MCF-7 Breast Cancer Cells.

As a potential means to reduce proliferation of breast cancer cells, a multiple-pathway approach with no effect on control cells was explored. The human interactome being constructed by the Center for Cancer Systems Biology will prove indispensable to understanding composite effects of multiple pathways, but its discovered protein-protein interactions require characterization. Accordingly, we explored the effects of regulators of one protein on downstream targets of the other protein. MCF-7 estrogen receptor-positive (ER+) breast cancer cells were treated with raloxifene to upregulate the TGF-β pathway and PX-866 to down-regulate the PI3K/Akt pathway. This resulted in highly significant downstream reduction of cell cycle proliferation in breast cancer cells with no significant proliferation reduction following similar treatment of noncancerous MCF10A breast epithelial cells. Reduced phosphorylation of p107 and substantial reduction of Rb phosphorylation were observed in response. The effects of reduced Rb and p107 phosphorylation were reflected in significant decline in E2F-1 transcriptional activity, which is dependent on pocket protein phosphorylation status. The reduced proliferation was related to decreased expression of cyclins, including E2F-1-regulated Cyclin E2, which was also in response to raloxifene and PX-866. All combinations of raloxifene and PX-866 produced significant or highly significant results for reduced MCF-7 cell proliferation, reduced Cyclin E2 transcription, and reduced Rb phosphorylation. These studies demonstrated that uncontrolled proliferation of ER+ breast cancer cells can be significantly reduced by combinational targeting of two relevant pathways. J. Cell. Biochem. 117: 1688-1696, 2016. © 2015 Wiley Periodicals, Inc.

Abstract B2-35: Efficient exploration of multi-cancer networks by generalized covariance selection and interactive web content

Abstract Statistical network modeling techniques are increasingly important tools to analyze cancer genomics data. However, current tools for network construction and interpretation are not designed to work across multiple diagnoses and technical platforms, thus limiting their applicability to comprehensive pan-cancer datasets such as the Cancer Genome Atlas (TCGA). Here, we describe a novel strategy to construct and analyze integrative network models heterogeneous data from multiple cancers. First, we introduce a generalization of sparse inverse covariance selection (SICS) designed to integrate genetic, epigenetic and transcriptional data from multiple cancers into a comparative network. The algorithm is shown to be statistically robust, effective at detecting direct pathway links in data from The Cancer Genome Atlas (TCGA), and uses a new strategy involving non-informative priors to balance different cancers and data types. Second, we propose to rationalize the interpretation of the derived networks by a new and publicly accessible tool (cancerlandscapes.org), in which derived models are explored as interactive web content, linked to several pathway and pharmacological databases. To evaluate the performance of the method, we constructed a model of genetic, epigenetic and transcriptional data for eight TCGA cancers, using data from 3900 patients. The derived model rediscovered known mechanisms and contained interesting predictions. Possible applications include the prediction of regulatory relationships between genes in particular cancers, comparison of network modules in across multiple forms of cancer, and identification of drug targets in relation to network structure. Citation Format: Teresia Kling, Patrik Johansson, Jose Sanchez, Voichita D. Marinescu, Rebecka Jornsten, Sven Nelander. Efficient exploration of multi-cancer networks by generalized covariance selection and interactive web content. [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 B2-35.

Abstract B2-27: Systems biological identification of optimal drug targets for combination therapy of non-small cell lung cancer

Abstract Non-small cell lung cancer (NSCLC) is the major subtype of lung cancer, accounting for about 85% of all lung patients. Recently, molecularly targeted inhibitors of receptor tyrosine kinase (RTK) such as epidermal growth factor receptor (EGFR) and fibroblast growth factor receptor (FGFR) signalling pathways have been highlighted as promising anti-tumor drugs because of the prevalence of genetic alterations within EGFR and FGFR signalling pathways in NSCLC patients. However, their functional effects have been reported as varying over a wide range depending on the patient-specific mutational conditions. To understand the underlying mechanism of the heterogeneous responses to EGFR and FGFR inhibitors and to find the optimal strategy for combination therapy, we have developed an integrative Boolean network model of EGFR, FGFR and DNA damage signalling pathways by integrating previous experimental evidences from the literature. Systematic perturbation analysis revealed that the multiple feedback regulations embedded in the signalling pathways play important roles in determining the responses to EGFR and FGFR inhibitors. We further analyzed the mechanism for the heterogeneous responses of cells to the EGFR and FGFR inhibitors under various mutational conditions and identified key genetic alterations that can lead to undesirable responses. Finally, extensive combinatorial perturbation analysis and subsequent comparison of the attractor profiles in the presence and absence of targeted inhibitors identified novel candidate drug targets for combinatorial therapy. Our analysis provides a new insight into the underlying mechanism of the heterogeneous cellular responses to EGFR and FGFR inhibitors and the optimal combinatorial treatment strategies. Acknowledgements: This work was supported by the National Research Foundation of Korea (NRF) grants funded by the Korea Government, the Ministry of Science, ICT & Future Planning (2014R1A2A1A10052404, 2013M3A9A7046303, and 2010-0017662) and by a grant of the Korean Health Technology R&D Project, Ministry of Health & Welfare, Republic of Korea. (HI13C2162). Citation Format: Jonghoon Lee, Kwang-Hyun Cho. Systems biological identification of optimal drug targets for combination therapy of non-small cell lung cancer. [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 B2-27.

Abstract IA24: Computational dissection of phenotypic and functional heterogeneity in cancer

Abstract Cells within a single tumor are known to display extensive phenotypic and functional heterogeneity. Many life-threatening features of cancer, including drug resistance, metastasis and relapse, are facets of intratumor heterogeneity. With emerging single-cell measurement technologies, the field is poised to make important strides in understanding and controlling this heterogeneity. However, these technologies require advances in analytical methods to interpret the complex data they produce. Using mass cytometry, which measures single cells in ~31 simultaneous proteomic features, we developed novel methods for analyzing phenotypic heterogeneity in cancer. We use AML as an example to demonstrate the power of our approach. The heart of our approach is Phenograph, a graph-based representation of single-cells which represents the phenotypic structure of the sample and can be partitioned into subsets of densely interconnected nodes, called communities. Using Phenograph, we deconstructed several AML samples into discrete phenotypes. Analyzing the resulting subpopulations provided insights into functional heterogeneity of AML. Phenograph can be applied to characterize heterogeneity and primitive subpopulations in additional cancers. Citation Format: Dana Pe'er. Computational dissection of phenotypic and functional heterogeneity in cancer. [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 IA24.

Abstract B2-59: PhyloSpan: Using multi-mutation reads to resolve subclonal architectures from heterogeneous tumor samples

Abstract We have developed a new method that uses high-throughput reads that span multiple somatic point mutations to reconstruct multiple, genetically diverse subclonal populations from one or more heterogeneous tumor samples. Tumors often contain multiple, genetically diverse subclonal populations, as predicted by the clonal theory of cancer. These subclonal populations develop through successive waves of expansion and selection and have differing abilities to metastasize and resist treatment. Identifying these sub-populations and their evolutionary relationships can help identify driver mutations associated with cancer development and progression. Subclonal reconstruction algorithms attempt to infer the prevalence and genotype of multiple, genetically-related subclonal populations using the variant allele frequency (VAF) of somatic variants. To date, these algorithms exclusively use data on individual somatic mutations. This restriction greatly reduces their ability to fully resolve phylogenic ambiguities. In some cases, it is possible to determine the mutation status of >1 mutation in a single cell, for example, when single reads cover multiple single nucleotide variants (SNVs). This type of information can add considerable power to the phylogenetic reconstruction of the tumor subclonal population. We have developed the PhyloSpan algorithm which attempts to infer the states of multiple SNVs in single cells, and then exploits that information in subclonal reconstruction. Our algorithm starts with phasing somatic SNVs by looking for reads / read-pairs that cover both a somatic mutation and germline heterozygous single nucleotide polymorphism (SNP). These germline SNPs are often available through profiling of normal tissue. PhyloSpan then identifies SNVs that are on the same chromosome and close enough to be covered by a single read or paired reads. These pairs of mutations provide more phylogenetic certainty than can be found by looking at mutations independently. For example, if those SNVs are found in the same evolutionary branch, then we expect to see some reads containing both mutations. If however, the SNVs are an separate branches then no reads should show both SNVs. PhyloSpan integrates this phylogenetic information, along with information about the VAF of each somatic SNV in order to perform subclonal reconstruction. Incorporating these various types of information, especially given the substantial uncertainty in phasing and NGS read content, requires a rigorous statistical approach and so we have developed a Bayesian non-parametric tree-based clustering algorithm, based on our existing PhyloWGS method. This algorithm not only infers the number of subclonal populations and their genotype but also provides a measure of uncertainty about this inference, enabling users to determine which parts of the subclonal reconstruction are certain and which parts remain ambiguous. While the number of SNVs a short-read length distance away from another SNV is small, a handful of such pairs are all that is needed to eliminate a substantial amount of ambiguity in subclonal reconstruction. Furthermore, long (>10k) read technologies, such as PacBio, can be used to supplement short read sequence. Our approach generalizes to permit the integration of single-cell sequencing with bulk tumor sequencing. Furthermore, we can also use our framework to identify a small number of SNVs for which low throughput assays would be most useful to resolve subclonal reconstruction ambiguity. We will present results applying our algorithm to whole genome sequencing data showing the added value of considering multiple SNVs compared to independent SNVs. Citation Format: Amit G. Deshwar, Levi Boyles, Jeff Wintersinger, Paul C. Boutros, Yee Whye Teh, Quaid Morris, Quaid Morris. PhyloSpan: Using multi-mutation reads to resolve subclonal architectures from heterogeneous tumor samples. [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 B2-59.

Abstract B1-05: R2: Accessible web-based genomics analysis and visualization platform for biomedical researchers

Abstract In this era of explosive high throughput (HT) genomics data generation, there is a growing need for accessible software solutions that can help unlock biological/clinical characteristics from such data. With the biomedical researcher (with limited or no bioinformatics skills) in mind, we developed a comprehensive web-based system called R2 (http://r2.amc.nl). R2 aids in the analysis and visualization of private/shielded as well as public high throughput data and their associated annotated features. The R2 platform consists of a database storing both publicly accessible as well as shielded datasets with unified gene annotation, supplemented with a large suite of tools that can be used on these data, and their associated annotation. As such the user experiences the same look & feel throughout the mining process. In the public section, R2 hosts over 60,000 HT samples. Next to gene expression (microarray and RNA-seq), the platform is also being employed in the integration, analysis and visualization of other data types, such as aCGH, SNP, ChIP, methylation, miRNA, and NGS whole genome sequencing information. R2 contains an expanding set of interactive analyses which are heavily inter-connected, allowing users to quickly hop from one view (representation of the data) to another. Analyses include, correlation, differential expression, gene sets, gene ontology, transcription factor binding sites, PCA, kmeans, Kaplan Meier scans, signature creation etc. Visualizations include, various gene oriented plots, heatmaps, circos, embedded genome browser, Venn, etc. Many parts of the R2 platform are publicly accessible through the http://r2.amc.nl portal. The R2 gene expression analysis tools have thus far been used in more than 180 peer-reviewed scientific publications including journals as Nature, Nature Genetics and Cell. R2 also supports restricted / shielded use and is used in many (inter)national collaborative efforts involving unpublished datasets. The webservers have been serving over 900.000 pages over the past 12 months (Nov 2014). Citation Format: Jan Koster, Jan J. Molenaar, Rogier Versteeg. R2: Accessible web-based genomics analysis and visualization platform for biomedical researchers. [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 B1-05.

Abstract B1-63: Prostate cancer classification using a transcriptome atlas

Abstract Background: Gene expression profiling has been used to define subclasses of cancers and in the assembly of potential prognostic signatures. Molecular categories are commonly defined by supervised or unsupervised cluster analysis of mRNA levels. However, despite efforts to profile prostate cancer using transcriptome data, the genetic alterations and biological processes that contribute to clinical progression have made disease subtype categorization difficult. In addition, most studies are limited in statistical power, and therefore do not accurately reflect the heterogeneity of the disease in the general population. The goal of this study was to generate molecular profiles of prostate cancer using an unprecedentedly large transcriptome dataset. Methods: We compiled a very large compendium of 50 human prostate cancer transcriptome datasets containing over 4,000 human prostate cancer specimens, which we named the “Prostate Cancer Transcriptome Atlas” (PCTA). From this resource, 38 datasets, containing over 2,000 specimens, were combined into a single, normalized and batch-corrected dataset using the median-centering method. Molecular categories were sought using 29 published gene expression signature profiles, including castration-resistance (CRPC), oncogene activation, AR activation, AR variants, EZH2, FOXA1, TMPRSS2–ERG fusion, stemness, PTEN inactivation, polycomb complex (PRC2) repression, cell proliferation, epithelial-mesenchymal transition, and proneural and neuroendocrine differentiation. Signature profiles were computed by the weighted Z-score method. Results: Unsupervised clustering identified subsets of tumors manifesting consistent molecular features across many specimens, including a category showing strong AR signatures coupled in concert with significantly repressed PRC2 signatures. These features were mostly seen in high Gleason score (>7) or metastatic prostate cancer. Another major group was an “inverse” signature, showing activated PRC2 and repressed AR signatures. This stratification was independently validated using two datasets: (1) from 545 formalin-fixed paraffin-embedded (FFPE) tissue samples from primary prostate cancer from the Mayo Clinic (PLoS One 2013; 8(6):e66855) and 281 prostate cancers from a watchful-waiting cohort recruited in Sweden (BMC Med Genomics 2010;3:8). This analysis also allowed us to observe other, distinct variant classes, including activation of AR-variants and EZH2, high enrichment of stemness, and proneural and neuroendocrine differentiation. Conclusion: These results show that analyzing gene signatures using an integrated collection of transcriptome profiles from multiple platforms, and thereby significantly increasing sample size, allows the assignment of provisional disease categories that are either difficult to observe or not detected even in large profiling studies. Citation Format: Sungyong You, Jayoung Kim, Michael R. Freeman. Prostate cancer classification using a transcriptome atlas. [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 B1-63.

Abstract PR09: Extracellular stiffness cues drive spatial reorganization of the genome to globally constrain RNA abundance

Abstract The stiffness of the extracellular matrix (ECM) drives mechanosignaling that regulates tissue development and malignancy. We previously showed that a stiff ECM disrupts tissue organization and enhances malignant progression by inducing cell invasion and migration. However, the specific transcriptional and molecular events in which mechanotransduction directs these phenotypes are not well understood. To clarify this process, we used a combination of genome-scale approaches to monitor changes in gene expression and protein abundance as a function of acinar morphogenesis and tissue homeostasis in three dimensional extracellular matrix hydrogels with tunable stiffness. Elevated ECM stiffness perturbed tissue homeostasis and reverted the transcriptional phenotype of differentiated mammary acini to resemble that observed in rapidly proliferating nonpolarized mammary cell aggregates. These findings suggest that tissue tension induces cellular changes that directly reflect higher-order tissue organizational states. We found that these changes involve the spatial rearrangement of peripheral chromatin, and that the expression levels of multiple histone deacetylases increase in organized tissues concurrently with elevated nuclear heterochromatin content, an effect that is abrogated in rigid ECM conditions. We support these observations by mapping mechanoresponsive peripheral heterochromatin elements via ChIPseq, enabling us to directly identify dynamic regions containing genes whose transcriptional activity is responsive to mechanical cues. Finally, using a combination of genomic, imaging, and molecular biology techniques we demonstrated that ECM compliance and tissue organization significantly influences global RNA abundance. Notably, this model presents formidable conceptual and practical challenges for the interpretation of genomic data. Collectively, this work indicates that tissue organization is critically dependent on the cellular mechanical environment, which qualitatively and quantitatively shapes the epigenetic and transcriptional landscape by mechanisms that have not yet been elucidated. Note: This abstract was not presented at the conference. Citation Format: Russell Bainer, Yoshihiro Yui, Shannon Mumenthaler, Parag Mallick, Lin Liu, Hua-Jun Wu, Ondrej Podlaha, Franziska Michor, Jan Liphardt, Jonathan Licht, Valerie Weaver. Extracellular stiffness cues drive spatial reorganization of the genome to globally constrain RNA abundance. [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 PR09.

Abstract B1-39: Multi-target drug combinations from single drug responses measured at the level of single cells using Mixture Nested Effects Models (MNEMs) applied to cancer

Abstract Purpose: To identify an optimal drug combination for addressing intratumor heterogeneity by measuring single drug responses at the level of single cells. Background: Drug combination strategies have been derived from drug-drug and drug-target interaction inferences made based on genomic network analysis, such as synthetic lethality network analysis and protein network analysis. A key limitation with these approaches is that they do not account for intratumor heterogeneity. A next generation high throughput single cell technology such as Cytometry Time-of-Flight (CyTOF) has the potential to circumvent this limitation by providing the key information for optimally targeting intratumor heterogeneity within the context of personalized medicine. Methods: We propose a systems level modeling framework for analyzing heterogeneous drug response data across a population of single cells. Our approach, called Mixture Nested Effects Models (MNEM), identifies multi-target drug combinations from single drug effects measured at the level of single cells. MNEM analyzes drug responses among subpopulations (aka clusters) of similar cells, where the subpopulations are defined by cellular surface markers. MNEM infers the nested drug target hierarchy (directed network) from a heterogeneous population of drug effects by combining subpopulation identification, subpopulation matching to infer the drug response pre- and post-treatment, and nested effects modeling for each subpopulation individually or through a mixture of subpopulations. An MNEM infers that Drug A targets are a superset of Drug B targets, if the heterogeneous downstream target effects resulting from Drug B are a noisy subset of those resulting from Drug A. This relationship is represented as a directed edge from Drug A to Drug B. From the graphical model of nested drug effects, a scoring function is superimposed to identify the optimal drug combinations. Results: We apply MNEM to CyTOF drug screening data derived from treating three well-established cancer cell lines: HeLa (ovarian), NCIH460 (lung) and HCT116 (colon) with Mek, mTOR, pP38MAPK, JnK I and PI3K inhibitors at different optimal dose levels in addition to the base line treatment TRAIL, an apoptotic stimulator. MNEM indicates that pP38MAPK and Mek inhibitors are important candidates for targeting the ovarian and lung cancer cell lines. The MNEM analysis from the colon cancer cell line data indicates that several optimal drug regimens may have equivalent efficacy. Survival data from drug combination invitro experiments using clonogenic assays support the results from MNEM. Overall, these analyses suggest that MNEM could be used to optimize drug combinations based on responses from single drugs measured at the single cell level. Conclusion: Intratumor heterogeneity in tumor response, measured at the level of single cells, upon exposure to single drugs may be combined through Mixture Nested Effects Modeling (MNEM) to infer optimal drug combinations. Citation Format: Benedict Anchang, Harris Fienberg, Sean Bendall, Robert Tibshirani, Sylvia K. Plevritis. Multi-target drug combinations from single drug responses measured at the level of single cells using Mixture Nested Effects Models (MNEMs) applied to cancer. [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 B1-39.

Abstract IA03: Towards the next clinical option: Experience from a precision cancer medicine trial

Abstract Understanding molecular mechanisms of response and resistance to anticancer therapies requires prospective patient follow-up and clinical and functional validation of both common and low frequency mutations. We describe an evidence-based precision medicine program designed to bring whole exome sequencing (WES) into clinical practice and illustrate how this can be utilized to identify novel predictive biomarkers associated with exceptional systemic response. Metastatic and treatment-resistant cancer patients are prospectively enrolled for paired metastatic tumor and normal WES. Using a comprehensive computational pipeline, point mutations, indels and copy number alterations are detected. Mutations are categorized as Category I-III based on actionability and a report is generated and discussed in tumor board. Patients are followed prospectively for correlation of molecular information with clinical response and patient outcomes. A rapid response functional team follows on key findings to expand actionability based on the WES results. In the first year, 154 tumor-normal pairs from 97 patients were sequenced, with an average coverage of 95X of over 21,000 genes and 16 somatic alterations detected on average per patient. In total, 16 mutations were Category I, 98 were Category II, and 1474 were Category III. Tumor purity ranged from 14 to near 100%. Among unexpected findings, a prostate cancer exceptional responder was identified that harbored a somatic hemizygous deletion of the DNA repair gene FANCA and likely loss of function of the second allele through germline missense variant. Follow-up experiments established that loss of FANCA function was associated with platinum hypersensitivity both in vitro and in patient-derived xenografts, thus providing biologic rationale and functional evidence for his extreme clinical response. The establishment of a clinical program for whole exome sequencing of metastatic tumors with prospective follow-up of patients can help identify candidate predictive biomarkers of response. Citation Format: Mark A. Rubin. Towards the next clinical option: Experience from a precision cancer medicine trial. [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 IA03.

Abstract B1-01: Whole genome expression based drug repurposing

Abstract Currently, personalizing cancer chemotherapy relies on pathology and more recently molecular biomarker-based approaches. However, as the driving biology is normally not fully understood, the majority of existing biomarkers do not capture a substantial proportion of variability in drug response. This partly explains the commonly observed lack of reproducibility of findings (e.g. from many conventional gene expression signatures) when these markers are applied to new datasets. We developed an approach to predict in vivo drug sensitivity that leverages whole-genome gene expression microarray data and allows the expression of every gene to influence the prediction by a small amount. Our approach builds statistical models from gene expression and drug sensitivity data in a very large panel of cell lines, then applies these models to gene expression data from primary tumor biopsies. In this study, we applied this approach to tumor samples collected in The Cancer Genome Atlas (TCGA). We derived predicted sensitivity for over one hundred drugs in each of the tumor samples. As a proof-of-concept, we demonstrated that a targeting agent (lapatinib) designed against a specific tumor marker (HER2 positive) is indeed predicted to be more sensitive in HER2 positive breast cancers when compared to the other type of cancers. Meanwhile, we identified other agents that exhibit similar or superior sensitivity when compared to commonly prescribed agents in different disease settings. These findings warrant further evaluation of these agents to be repurposed for possible new indications. Interestingly, some of our derived drug sensitivity predictive estimates are correlated with observed survival outcomes in certain cancer patients. When screening all tumor types based on their molecular profiles, we defined several classes of drugs that maybe differentially effective based on tumor molecular profiles. In conclusion, a genome-wide expression drug sensitivity model built in cell lines can be a powerful approach in repurposing drug in cancer treatment. Citation Format: Paul Geeleher, Steven Bhutra, Jacqueline Wang, R. Stephanie Huang. Whole genome expression based drug repurposing. [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 B1-01.

Abstract B2-55: Critical role of immune spatial heterogeneity and the molecular scaffold in estrogen receptor-negative breast cancer

Abstract Purpose: The abundance of tumor-infiltrating lymphocytes has been associated with a favorable prognosis in estrogen receptor-negative breast cancer. However, a high degree of spatial heterogeneity in lymphocytic infiltration is often observed in histology samples and its clinical implications and underpinning molecular scaffold remain unclear. Materials and methods: To quantify spatial heterogeneity in the distribution of tumor infiltrating lymphocytes, we combined automated histological image processing with methods of spatial statistics used in ecological data analysis. Hematoxylin and eosin-stained sections from two cohorts of estrogen receptor-negative breast cancer patients (discovery: n=120; validation: n=125) were processed with our automated cell classification algorithm to identify the location of lymphocytes and cancer cells. Subsequently, hotspot analysis (Getis-Ord Gi*) was applied to identify statistically significant hotspots of cancer and immune cells, defined as tumor regions with a significantly high number of cancer or immune cells respectively. To identify molecular aberrations that explain tumor spatial heterogeneity, we integrated our image-based hotspots results with microarray gene expression and copy number data profiled for the same set of tumors. Molecular data were generated with tumor materials sandwiched between these sections, thereby maximizing the biological relevance of multiple data types being generated. Results: We found that the amount of colocalized cancer and immune hotspots weighted by tumor area, rather than number of cancer or immune hotspots, significantly correlates with a better prognosis in estrogen receptor-negative breast cancer in uni- and multivariate Cox analysis. Moreover, colocalization of cancer and immune hotspots further stratified patients with immune cell-rich tumors. Subsequently, we developed a bioinformatics tool, iMmune hOTspoT Omics (iMOTTO), to explain the hotspots as a clinically relevant phenotype using molecular profiling data. Our preliminary analysis revealed significant correlations between this phenotype and expression of immune-specific genes such as CD79, CCL19 and SLAMF1, as well as an immunotherapy target, CTLA4. By incorporating the expression of hotspots-associated genes and copy number alteration data into a multivariate regression model, we aim to define a minimal set of genes to explain the observed degree of cancer-immune hotspot colocalization. Conclusion: Taken together, our study demonstrates the importance of quantifying not only the abundance of lymphocytes but also their spatial variation in the tumor specimen for which methods from other disciplines such as spatial statistics can be successfully applied. Systematic integration of histology and omics data revealed key immune regulators as well as novel genes which warrant further investigation to help elucidate the biological processes underlying immune spatial heterogeneity with potentially important clinical implications. Furthermore, our computational approach can be adapted for studying other cancer types for which immunotherapy has been applied, such as melanoma and non-small cell lung cancer, where our hotspot measures can potentially serve as prognostic and predictive biomarkers. Citation Format: Sidra Nawaz, Andreas Heindl, Andrea Agostinelli, Yinyin Yuan. Critical role of immune spatial heterogeneity and the molecular scaffold in estrogen receptor-negative breast cancer. [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 B2-55.

Abstract B1-56: Distinct roles of copy number and loss-of-heterozygosity in predicting prognosis for breast cancer patients

Abstract Purpose: Novel molecular stratification of the breast cancer population based on genomic aberrations has been recently reported.[1] The main focus has been on deriving prognostic classifiers by integrating gene expression and copy number aberrations (CNAs) data, but the role of other events, such as loss-of-heterozygosity (LOH), has not been extensively characterized in this integrated landscape. Here we consider the respective role of CNAs and LOH in breast cancer, in the context of an integrated genomic analysis, and their impact on clinical outcome. Methods: Currently, retrospective data with gene expression, CNA and LOH were used from 219 patients with early primary breast cancer who were resected in the period 1989-2003.[2] CNA and LOH calls were made using OncoSNP v1.4, a Bayesian model driven genomic analysis tool to call genomic aberrations and characterize LOH status, and statistical analyses were performed in R v3.0.1. The evaluated clinical outcome was distant-recurrence free survival (DRFS). Two gene sets were tested: A) general cancer drivers from the COSMIC cancer gene census and B) all genes across the genome. Significantly amplified (increased DNA copy number with respect to normal tissue), copy neutral (no change in copy number) and deleted cancer drivers were identified by deriving null distributions from permutation analysis. Survival analysis was done using both LogRank and Cox Proportional Hazards methods. The effect of LOH on the gene expression was tested using Mann-Whitney. Benjamini and Hochberg method was used to adjust for multiple testing. Results: A higher than expected LOH occurrence was detected in 168 out of 523 COSMIC defined cancer drivers. Of these, 10 were significantly amplified, 94 were copy neutral and 64 deleted. Within these three copy number groups, patients with LOH vs no LOH were compared for DRFS. LOH was a prognostic event in 14 out of 94 copy neutral group of genes, while 4 and 2 genes were found in the amplified and deleted genes respectively. Furthermore, tumors showing copy neutral LOH events had worse outcome than tumors where these genes are amplified (no LOH) or deleted (LOH). A whole genome analysis showed that a higher global LOH frequency, across the entire genome, is associated with poor prognosis as measured by a decreased DRFS, irrespective of the global level of CNAs in the tumour. Furthermore, 386 of the 5184 genes that showed a higher than expected LOH occurrence, were significantly associated with DRFS with respect to LOH status, and 51 of these were so independently of CNAs and global LOH. Conclusions: LOH in copy neutral patients has an effect on clinical outcome in a subset of cancer drivers and overall genes, irrespective of the global genomic alteration status. This finding supports an approach of measuring and integrating multiple genomic data information, which expectantly improves the accuracy of identifying subgroups of patients, resulting in improved treatment selection. Ongoing work, which will be presented at the meeting, is focused on characterizing these events, and confirming these analyses in the large TCGA dataset[3] with multiple cancer types.

Abstract PR13: Systems and structural biology approaches to elucidate new effectors in KRAS mutant tumors

Abstract Activated KRAS interacts directly and indirectly with numerous downstream effector pathways that are still elusive or poorly characterized. Furthermore, it has become clear over the past few years that the KRAS regulatory gene network (interactome) is highly cell-context dependent and thus tumor specific. We sought to computationally elucidate novel KRAS effectors, modulators, and interactors in Lung Adenocarcinomas (LUAD) and Colon Adenocarcinomas (COAD) by integrating computational and biochemical reverse engineering methods. Specifically, results from ARACNe and MINDy based analysis of KRAS networks were combined with Protein-protein interaction (PPI) predictions based on the PrePPI algorithm. These algorithms detect, respectively, transcriptional, post-translational, and protein-protein interactions respectively. The MARINa/VIPER algorithm was used to analyze the networks resulting from these analyses to identify key modulators and effectors of KRAS activity, including modulators harboring genetic alterations that phenocopy aberrant KRAS activity. In addition, we collaborated with the Jackson (Stanford) and Vidal (Harvard) labs to include high-throughput tandem affinity mass spectrometry data and yeast 2 hybrid screens in these analyses. Validation of computationally inferred effectors, using pooled shRNA silencing assays in primary lung cancer cells isolated from a KrasLSL/p53f/f mouse model of LUAD validated novel genes as likely KRASmut effectors and putative synthetic lethal genes. At least two hairpins against RNF43, E4F1, HMCN1, MYO9A, RHOT2, SETD1B and MAP3K2 scored in this assay. These results suggest that KRAS network analyses using computational methods can identify novel members of the KRAS interactome. Citation Format: Joshua Broyde, David Simpson, David A. Wah, Federico M. Giorgi, Donald Petrey, Mariano J. Alvarez, Antonina Silkov, Alexander Lachmann, David E. Hill, Marc Vidal, Peter Jackson, Barry Honig, Alejandro Sweet-Cordero, Andrea Califano. Systems and structural biology approaches to elucidate new effectors in KRAS mutant tumors. [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 PR13.

Abstract B1-26: NetFill: A network-based method to identify a parsimonious set of mediators to link mRNA and protein signatures using existing knowledge

Abstract Introduction: Despite our general understanding of the genetic information flow from DNA to RNA to protein, the integrated analysis of multi-level omics data in a biologically meaningful way remains a challenging task due to the complex and dynamic nature of disease associated changes in molecular regulatory mechanisms. A case in point is the often observed phenomenon that key signatures associated with the same phenotypic difference do not share much in common between the mRNA and proteomic data even when they are measured on a common set of samples. Methods: We developed a computational method, NetFill, that integrates prior knowledge into the analysis of expression data to identify a parsimonious set of molecular mediators to connect the mRNA and protein signatures with plausible pathways/networks. NetFill formulates the problem mathematically as a flow-based linear programming under constraints to identify mediator/proteins in the protein-protein interactome network with the minimal cost cumulated from selected nodes and selected edges to link the mRNA and protein signatures. Data dependent node score and edge score were computed from the mRNA and proteomic expression data using Bayes' rule. By assuming conditional independence between different types of evidence, the node score was derived from the p values of t-tests between two conditions. Edge score was derived from the p-value of correlation between two proteins and edge confidence score from static PPI network. In the current implementation of NetFill, nine protein-protein interaction databases (BIND, BioGRID, DIP, HPRD, IntAct, MINT, MIPS, PDZBase, and Ractome) and three pathway databases (KEGG, Biocarta, and NCI) were combined to generate a large protein-protein interactome network with 101,398 interactions. To assess the confidence score of selected interactions, the network search algorithm was applied on 100 bootstrap with replacement samples of the mRNA and protein data. The final output of NetFill was a high-confidence network in which all edges had occurred in > 50% of the bootstrap sampling generated networks. Results: Resulted from separate consensus clustering analyses, an mRNA signature of 213 genes and a proteomic signature of 73 proteins for a common subtype/cluster of ovarian high-grade serous carcinoma (HGSC) samples had only 40 overlapping gene/proteins. NetFill identified a parsimonious network with 142 additional mediators to link the mRNA and proteomic signatures. Functional annotation revealed that Wnt signaling pathway (p = 0.01) and MAPK signaling pathway (p = 0.03) were significantly enriched in the identified network, indicating that these pathways could be potentially involved in this subtype of ovarian HGSC. Conclusion: The NetFill method and reported example demonstrated that existing knowledge of plausible interactions among genes and proteins could be used to help to functionally integrate mRNA and protein data in multi-level omics characterization of samples. Citation Format: Li Chen, Bai Zhang, Hui Zhang, Yue Wang, Daniel W. Chan, Zhen Zhang. NetFill: A network-based method to identify a parsimonious set of mediators to link mRNA and protein signatures using existing knowledge. [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 B1-26.