Abstract 432: BIN3 is a novel 8p21 tumor suppressor gene that regulates the attachment checkpoint in epithelial cells

An important characteristic of multicellular organisms is the control that the tissue architecture exerts on the fate of individual cells. Epithelial cells sense their location through interactions with the extracellular matrix (ECM) and remove themselves by programmed cell death (anoikis) when those interactions are disturbed. Importantly, anoikis is a line of defense that has to be circumvented by cancerous epithelial cells to be able to leave their home environment and establish long distance metastases. Here, by combining a genome-wide RNAi screen and a novel algorithm to study copy number alterations (ISAR-DEL), we identify the Bridging Integrator-3 (BIN3) as a novel 8p21 tumor suppressor gene whose inactivation promotes escape from anoikis in epithelial cancers. Mechanistically, we link the tumor suppression function of BIN3 to its ability to relocate to the cell membrane after cell detachment and to induce a proapoptotic cascade. This death signaling is mediated by CDC42 activation of the P38-α stress pathway and the consequent accumulation of the apoptotic facilitator BimEL. Our results identify BIN3 as a novel epithelial tumor suppressor gene, provide novel insights on the mechanisms of attachment tumor suppressor checkpoint and highlight the importance of anoikis escape in driving cell transformation and metastasis in human cancer.

Note: This abstract was not presented at the meeting.

Citation Format: Netonia Marshall, Felix Sanchez, David Llobet, Ruth Rodriguez Barrueco, Veronica Castro, Dylan Kotlia, Maira Pires, Patricia Villagrasa, Preeti Putcha, Ramon Parsons, Dana Pe'er, Jose Silva. BIN3 is a novel 8p21 tumor suppressor gene that regulates the attachment checkpoint in epithelial cells. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 432. doi:10.1158/1538-7445.AM2014-432

Single-cell trajectory detection uncovers progression and regulatory coordination in human B cell development

Tissue regeneration is an orchestrated progression of cells from an immature state to a mature one, conventionally represented as distinctive cell subsets. A continuum of transitional cell states exists between these discrete stages. We combine the depth of single-cell mass cytometry and an algorithm developed to leverage this continuum by aligning single cells of a given lineage onto a unified trajectory that accurately predicts the developmental path de novo. Applied to human B cell lymphopoiesis, the algorithm (termed Wanderlust) constructed trajectories spanning from hematopoietic stem cells through to naive B cells. This trajectory revealed nascent fractions of B cell progenitors and aligned them with developmentally cued regulatory signaling including IL-7/STAT5 and cellular events such as immunoglobulin rearrangement, highlighting checkpoints across which regulatory signals are rewired paralleling changes in cellular state. This study provides a comprehensive analysis of human B lymphopoiesis, laying a foundation to apply this approach to other tissues and "corrupted" developmental processes including cancer.

The onset of p53 loss of heterozygosity is differentially induced in various stem cell types and may involve the loss of either allele

p53 loss of heterozygosity (p53LOH) is frequently observed in Li-Fraumeni syndrome (LFS) patients who carry a mutant (Mut) p53 germ-line mutation. Here, we focused on elucidating the link between p53LOH and tumor development in stem cells (SCs). Although adult mesenchymal stem cells (MSCs) robustly underwent p53LOH, p53LOH in induced embryonic pluripotent stem cells (iPSCs) was significantly attenuated. Only SCs that underwent p53LOH induced malignant tumors in mice. These results may explain why LFS patients develop normally, yet acquire tumors in adulthood. Surprisingly, an analysis of single-cell sub-clones of iPSCs, MSCs and ex vivo bone marrow (BM) progenitors revealed that p53LOH is a bi-directional process, which may result in either the loss of wild-type (WT) or Mut p53 allele. Interestingly, most BM progenitors underwent Mutp53LOH. Our results suggest that the bi-directional p53LOH process may function as a cell-fate checkpoint. The loss of Mutp53 may be regarded as a DNA repair event leading to genome stability. Indeed, gene expression analysis of the p53LOH process revealed upregulation of a specific chromatin remodeler and a burst of DNA repair genes. However, in the case of loss of WTp53, cells are endowed with uncontrolled growth that promotes cancer.

Single-cell trajectory detection uncovers progression and regulatory coordination in human B cell development

Tissue regeneration is an orchestrated progression of cells from an immature state to a mature one, conventionally represented as distinctive cell subsets. A continuum of transitional cell states exists between these discrete stages. We combine the depth of single-cell mass cytometry and an algorithm developed to leverage this continuum by aligning single cells of a given lineage onto a unified trajectory that accurately predicts the developmental path de novo. Applied to human B cell lymphopoiesis, the algorithm (termed Wanderlust) constructed trajectories spanning from hematopoietic stem cells through to naive B cells. This trajectory revealed nascent fractions of B cell progenitors and aligned them with developmentally cued regulatory signaling including IL-7/STAT5 and cellular events such as immunoglobulin rearrangement, highlighting checkpoints across which regulatory signals are rewired paralleling changes in cellular state. This study provides a comprehensive analysis of human B lymphopoiesis, laying a foundation to apply this approach to other tissues and "corrupted" developmental processes including cancer.

The miR-424(322)/503 cluster orchestrates remodeling of the epithelium in the involuting mammary gland

The mammary gland undergoes continuous remodeling. Llobet-Navas et al. identify the microRNA cluster miR-424(322)/503 as an important regulator of epithelial involution after pregnancy. TGF-β regulates the expression of this miR cluster, which in turn targets BCL-2 and IGF1R. This work suggests a model in which activation of the TGF-β pathway after weaning induces the transcription of the miR-424(322)/503 cluster to down-regulate the expression of key genes.

The mammary gland is a very dynamic organ that undergoes continuous remodeling. The critical regulators of this process are not fully understood. Here we identify the microRNA cluster miR-424(322)/503 as an important regulator of epithelial involution after pregnancy. Through the generation of a knockout mouse model, we found that regression of the secretory acini of the mammary gland was compromised in the absence of miR-424(322)/503. Mechanistically, we show that miR-424(322)/503 orchestrates cell life and death decisions by targeting BCL-2 and IGF1R (insulin growth factor-1 receptor). Furthermore, we demonstrate that the expression of this microRNA cluster is regulated by TGF-β, a well-characterized regulator of mammary involution. Overall, our data suggest a model in which activation of the TGF-β pathway after weaning induces the transcription of miR-424(322)/503, which in turn down-regulates the expression of key genes. Here, we unveil a previously unknown, multilayered regulation of epithelial tissue remodeling coordinated by the microRNA cluster miR-424(322)/503.

Genotype-environment interactions reveal causal pathways that mediate genetic effects on phenotype

Unraveling the molecular processes that lead from genotype to phenotype is crucial for the understanding and effective treatment of genetic diseases. Knowledge of the causative genetic defect most often does not enable treatment; therefore, causal intermediates between genotype and phenotype constitute valuable candidates for molecular intervention points that can be therapeutically targeted. Mapping genetic determinants of gene expression levels (also known as expression quantitative trait loci or eQTL studies) is frequently used for this purpose, yet distinguishing causation from correlation remains a significant challenge. Here, we address this challenge using extensive, multi-environment gene expression and fitness profiling of hundreds of genetically diverse yeast strains, in order to identify truly causal intermediate genes that condition fitness in a given environment. Using functional genomics assays, we show that the predictive power of eQTL studies for inferring causal intermediate genes is poor unless performed across multiple environments. Surprisingly, although the effects of genotype on fitness depended strongly on environment, causal intermediates could be most reliably predicted from genetic effects on expression present in all environments. Our results indicate a mechanism explaining this apparent paradox, whereby immediate molecular consequences of genetic variation are shared across environments, and environment-dependent phenotypic effects result from downstream integration of environmental signals. We developed a statistical model to predict causal intermediates that leverages this insight, yielding over 400 transcripts, for the majority of which we experimentally validated their role in conditioning fitness. Our findings have implications for the design and analysis of clinical omics studies aimed at discovering personalized targets for molecular intervention, suggesting that inferring causation in a single cellular context can benefit from molecular profiling in multiple contexts.

Mapping differentiation under mixed culture conditions reveals a tunable continuum of T cell fates

An experimental and theoretical study of T cell differentiation in response to mixed-input conditions reveals that cells can tune between Th1 and Th2 states through a continuum of mixed phenotypes.

Cell differentiation is typically directed by external signals that drive opposing regulatory pathways. Studying differentiation under polarizing conditions, with only one input signal provided, is limited in its ability to resolve the logic of interactions between opposing pathways. Dissection of this logic can be facilitated by mapping the system's response to mixtures of input signals, which are expected to occur in vivo, where cells are simultaneously exposed to various signals with potentially opposing effects. Here, we systematically map the response of naïve T cells to mixtures of signals driving differentiation into the Th1 and Th2 lineages. We characterize cell state at the single cell level by measuring levels of the two lineage-specific transcription factors (T-bet and GATA3) and two lineage characteristic cytokines (IFN-γ and IL-4) that are driven by these transcription regulators. We find a continuum of mixed phenotypes in which individual cells co-express the two lineage-specific master regulators at levels that gradually depend on levels of the two input signals. Using mathematical modeling we show that such tunable mixed phenotype arises if autoregulatory positive feedback loops in the gene network regulating this process are gradual and dominant over cross-pathway inhibition. We also find that expression of the lineage-specific cytokines follows two independent stochastic processes that are biased by expression levels of the master regulators. Thus, cytokine expression is highly heterogeneous under mixed conditions, with subpopulations of cells expressing only IFN-γ, only IL-4, both cytokines, or neither. The fraction of cells in each of these subpopulations changes gradually with input conditions, reproducing the continuous internal state at the cell population level. These results suggest a differentiation scheme in which cells reflect uncertainty through a continuously tuneable mixed phenotype combined with a biased stochastic decision rather than a binary phenotype with a deterministic decision.

During cell differentiation, progenitor cells respond to external signals that drive the expression of genes that are characteristic of the differentiated cell states. This process is controlled by gene regulatory networks that typically involve positive autoregulation and cross-inhibition between master regulators of the two differentiated states. Mapping the system's response to mixtures of external signals can help us to understand the operational logic of these binary cell fate decisions. Here, we study differentiation of CD4⁺ T cells into Th1 and Th2 lineages under mixed-input conditions, at the single cell level. We reveal that cell state is not restricted to a small number of well-defined phenotypes, but rather tunes through a continuum of mixed-phenotype states in which levels of lineage-specifying transcription factors gradually change with the levels of the two inputs. Using mathematical modeling we establish the conditions under which the system has one stable steady state that continuously tunes in response to changes in levels of the inputs. Results of this model qualitatively explain our experimental observations. We further characterize expression patterns of downstream lineage-specific genes—cytokines that are driven by the two master regulators upon cell re-stimulation. We find a highly heterogeneous population with cells expressing either one of the cytokines, both cytokines, or neither. Of note, the fraction of cells in these subpopulations continuously tunes with input levels, thus reproducing a tunable state at the cell population level. Our results can be explained by a two-stage scheme in which the gene regulatory network is responsible for a continuously tunable cell state, which is translated into a heterogeneous cytokine expression pattern through uncorrelated and biased stochastic processes.

RHPN2 drives mesenchymal transformation in malignant glioma by triggering RhoA activation

Mesenchymal transformation is a hallmark of aggressive glioblastoma (GBM). Here, we report the development of an unbiased method for computational integration of copy number variation, expression, and mutation data from large datasets. Using this method, we identified rhophilin 2 (RHPN2) as a central genetic determinant of the mesenchymal phenotype of human GBM. Notably, amplification of the human RHPN2 gene on chromosome 19 correlates with a dramatic decrease in the survival of patients with glioma. Ectopic expression of RHPN2 in neural stem cells and astrocytes triggered the expression of mesenchymal genes and promoted an invasive phenotype without impacting cell proliferation. Mechanistically, these effects were implemented through RHPN2-mediated activation of RhoA, a master regulator of cell migration and invasion. Our results define RHPN2 amplification as a central genetic determinant of a highly aggressive phenotype that directs the worst clinical outcomes in patients with GBM.

Abstract 3168: Helios identifies novel oncogenes in breast cancer by integrating genomic characterization of primary tumors and functional shRNA-screens

Genomic profiling of tumors has uncovered a staggering diversity of recurrent aberrations. However, inferring functionally important driver genes from this data remains difficult_particularly in the case of copy-number aberrations (CNAs) that often span many genes. Genome-wide functional shRNA screens have been a useful orthogonal approach for discovering drivers. The integration of observational data from primary tumors with functional data on cell lines provides an unprecedented opportunity for the identification of driver genes. Unfortunately, most current analysis is limited to naïve intersection of top scoring candidates and thus more powerful computational methods are needed. We have developed Helios, a novel Bayesian algorithm that integrates genomic data from primary tumors with functional shRNA screens in cell lines, gaining unprecedented sensitivity and specificity in identifying drivers. Applying Helios to TCGA breast cancer data led to the recapitulation of many known oncogenes as well as to the identification and validation of two novel oncogenes involved in chromatin regulation. Importantly, many of the drivers pinpointed by Helios were not identified on the basis of any one data type alone.

Helios uses shRNA data in a novel fashion by employing a new score measuring oncogene addiction, a phenotype associated with many key cancer drivers. It integrates this with CNA, sequence mutation, and RNA expression data into a single probabilistic score for each gene which is then used to assess the most-likely driver gene in a region of recurrent CNA.

We applied Helios to TCGA breast cancer data paired with two independent genome-wide shRNA screens on breast cancer cell lines. This identified many previously known oncogenes including FOXA1, ERBB2, PIK3CA, CCND1, IGF1R, BCL2, CDK4, ESR1, MYC, EGFR, GAB1, CCNE1, FGFR2, FGFR3, MYC as the top-scoring candidates in their respective amplified regions. We validated a number of novel predictions in vitro and selected two candidate oncogenes involved in chromatin remodeling for in depth follow up. The contribution of both genes to cancer was confirmed in vitro and in vivo, enhancing colony formation in agar and increasing tumor size in mouse models. One novel oncogene showed evidence of association with invasion and metastasis in a lung-cancer model. Another novel oncogene resides in a frequently amplified region in several epithelial cancers such as lung, bladder, stomach and ovarian carcinomas. Taken together, we have demonstrated that Helios is a powerful “in-silico” screen that can accelerate discovery of driver mutations in cancer.

Citation Format: Felix Sanchez Garcia, Patricia Villagrasa, Junji Matsui, Bo-Juen Chen, Dylan Kotliar, Veronica Castro, Jose M. Silva, Dana Pe'er. Helios identifies novel oncogenes in breast cancer by integrating genomic characterization of primary tumors and functional shRNA-screens. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 3168. doi:10.1158/1538-7445.AM2013-3168

Normalization of mass cytometry data with bead standards

Mass cytometry uses atomic mass spectrometry combined with isotopically pure reporter elements to currently measure as many as 40 parameters per single cell. As with any quantitative technology, there is a fundamental need for quality assurance and normalization protocols. In the case of mass cytometry, the signal variation over time due to changes in instrument performance combined with intervals between scheduled maintenance must be accounted for and then normalized. Here, samples were mixed with polystyrene beads embedded with metal lanthanides, allowing monitoring of mass cytometry instrument performance over multiple days of data acquisition. The protocol described here includes simultaneous measurements of beads and cells on the mass cytometer, subsequent extraction of the bead-based signature, and the application of an algorithm enabling correction of both short- and long-term signal fluctuations. The variation in the intensity of the beads that remains after normalization may also be used to determine data quality. Application of the algorithm to a one-month longitudinal analysis of a human peripheral blood sample reduced the range of median signal fluctuation from 4.9-fold to 1.3-fold.

Abstract 3958: Multi-Reg: An integrative parallel approach to uncover drivers of cancer

Tumor samples harbor a vast number of genomic alterations of various kinds, and it is difficult to distinguish driver alterations that contribute to oncogenesis from passenger alterations. Most computational methods identify driver genes by focusing on the most frequent copy number alterations. Previous work in our lab has led to the development of CONEXIC, a Bayesian algorithm that identifies candidate driver genes in cancer and links them to gene expression signatures they govern by integrating copy number and gene expression (Akavia et al, Cell, 2010). This algorithm was applied to data from melanoma cell lines, where it correctly identified known drivers, such as MITF and KLF6, and connected them to their known targets. In addition, it predicted novel tumor dependencies not previously implicated in melanoma, which were validated experimentally. Drivers may act concurrently, where not only the strongest one is important. For example, either PTEN deletion or AKT activation can lead to a similar expression signature and phenotype. Therefore, we developed a new algorithm, based on the same principles as CONEXIC, with multiple improvements. The new algorithm - Multi-Reg - is capable of detecting multiple candidate regulators that can all act in parallel to regulate an expression signature, and is also capable of integrating mutations in addition to copy number and gene expression data. We applied Multi-Reg to glioblastoma data from The Cancer Genome Atlas (TCGA). This data includes copy number, gene expression and mutations for hundreds of primary tumor samples. Multi-Reg has identified putative drivers that were missed by CONEXIC, such as the known genes PDGFRA and NF1, in addition to EGFR & ERB2 (identified by CONEXIC). Additionally, since Multi-Reg candidate drivers act in parallel, we can group them by shared targets. For example, EGFR & ERRB2 induce the same genes, which represent the Mesenchymal subtype of glibolastoma. These same genes are repressed by PDGFRA & NF1. This matches the known behavior of glioblastoma subtypes, where EGFR & PDGFRA characterize the Mesenchymal and Proneural subtypes, respectively. Thus, our results correctly identify known drivers of glioblastoma. Additionally, Multi-Reg results identified RHPN2 as a novel oncogenic factor controlling a gene expression signature related to adhesion. Validation has shown that while RHPN2 has no effect on cell proliferation it induces invasiveness in glioblastoma cell lines. This shows that Multi-Reg cannot only discover drivers, but can link them to the oncogenic phenomena they controls, and suggest the appropriate biological validation for it.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 3958. doi:1538-7445.AM2012-3958

Inference of modules associated to eQTLs

Cataloging the association of transcripts to genetic variants in recent years holds the promise for functional dissection of regulatory structure of human transcription. Here, we present a novel approach, which aims at elucidating the joint relationships between transcripts and single-nucleotide polymorphisms (SNPs). This entails detection and analysis of modules of transcripts, each weakly associated to a single genetic variant, together exposing a high-confidence association signal between the module and this 'main' SNP. To explore how transcripts in a module are related to causative loci for that module, we represent such dependencies by a graphical model. We applied our method to the existing data on genetics of gene expression in the liver. The modules are significantly more, larger and denser than found in permuted data. Quantification of the confidence in a module as a likelihood score, allows us to detect transcripts that do not reach genome-wide significance level. Topological analysis of each module identifies novel insights regarding the flow of causality between the main SNP and transcripts. We observe similar annotations of modules from two sources of information: the enrichment of a module in gene subsets and locus annotation of the genetic variants. This and further phenotypic analysis provide a validation for our methodology.

Using systems and structure biology tools to dissect cellular phenotypes

The Center for the Multiscale Analysis of Genetic Networks (MAGNet, http://magnet.c2b2.columbia.edu) was established in 2005, with the mission of providing the biomedical research community with Structural and Systems Biology algorithms and software tools for the dissection of molecular interactions and for the interaction-based elucidation of cellular phenotypes. Over the last 7 years, MAGNet investigators have developed many novel analysis methodologies, which have led to important biological discoveries, including understanding the role of the DNA shape in protein-DNA binding specificity and the discovery of genes causally related to the presentation of malignant phenotypes, including lymphoma, glioma, and melanoma. Software tools implementing these methodologies have been broadly adopted by the research community and are made freely available through geWorkbench, the Center's integrated analysis platform. Additionally, MAGNet has been instrumental in organizing and developing key conferences and meetings focused on the emerging field of systems biology and regulatory genomics, with special focus on cancer-related research.

Abstract A26: Multi-Reg: An integrative parallel approach to uncover drivers of cancer

Tumor samples harbor a vast number of genomic alterations of various kinds, and it is not easy to distinguish driver alterations that contribute to oncogenesis from passenger alterations. Most computational methods attempt to identify driver alternations by focusing on the most frequent alterations. Previous work in our lab has led to the development of CONEXIC, a Bayesian framework for integrating copy number and gene expression to identify candidate driver genes in cancer and to link them to gene expression signatures they regulate (Akavia et al, Cell, 2010). This framework was applied to data from melanoma cell lines, where it correctly identified known drivers (MITF) and connected them to their known targets. In addition, it predicted novel tumor dependencies not previously implicated in melanoma, which were confirmed experimentally.

In general, current algorithms identify only one driver (the strongest) controlling a gene expression signature. However, drivers may act in parallel, where not only the strongest one is important. For example, either PTEN deletion or AKT activation can lead to a similar expression signature and phenotype. Therefore, we developed a new algorithm, based on the same principles as CONEXIC with multiple improvements. The new algorithm (called Multi-Reg) is capable of detecting multiple candidate regulators that can all act in parallel to regulate an expression signature. This algorithm can integrate mutations in addition to copy number and expression. Finally, we have designed Multi-Reg to be easier and quicker to run and more robust.

We applied Multi-Reg to glioblastoma data from The Cancer Genome Atlas (TCGA). This data includes copy number, gene expression and mutations for hundreds of primary tumor samples. We found 84 candidate regulators that were missed by CONEXIC, but discovered by Multi-Reg. Because of Multi-Reg's ability to search for genes working in parallel, it identified FGFR3, PDGFRA and NF1, in addition to EGFR & MET (identified by CONEXIC), as important candidate drivers. Additionally, Multi-Reg results identified RHPN2 as a novel oncogenic factor controlling a gene expression signature related to invasion and migration. Validation has shown that RHPN2 has limited effect on cell proliferation but induces invasiveness in glioblastoma cell lines.

Our results correctly identify known drivers of glioblastoma progression, including oncogenes such as EGFR, MET, CEBPB and tumor suppressors such as p16 and NF1. Multi-Reg also has the capability to identify more regulators than CONEXIC, and has correctly linked RHPN2 to the oncogenic phenomena it controls.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the Second AACR International Conference on Frontiers in Basic Cancer Research; 2011 Sep 14-18; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2011;71(18 Suppl):Abstract nr A26.

Single-cell mass cytometry of differential immune and drug responses across a human hematopoietic continuum

Flow cytometry is an essential tool for dissecting the functional complexity of hematopoiesis. We used single-cell "mass cytometry" to examine healthy human bone marrow, measuring 34 parameters simultaneously in single cells (binding of 31 antibodies, viability, DNA content, and relative cell size). The signaling behavior of cell subsets spanning a defined hematopoietic hierarchy was monitored with 18 simultaneous markers of functional signaling states perturbed by a set of ex vivo stimuli and inhibitors. The data set allowed for an algorithmically driven assembly of related cell types defined by surface antigen expression, providing a superimposable map of cell signaling responses in combination with drug inhibition. Visualized in this manner, the analysis revealed previously unappreciated instances of both precise signaling responses that were bounded within conventionally defined cell subsets and more continuous phosphorylation responses that crossed cell population boundaries in unexpected manners yet tracked closely with cellular phenotype. Collectively, such single-cell analyses provide system-wide views of immune signaling in healthy human hematopoiesis, against which drug action and disease can be compared for mechanistic studies and pharmacologic intervention.

Abstract SY17-02: A systems approach to understanding tumor specific drug response

Systematic characterization of cancer genomes has revealed a staggering complexity of aberrations among individuals, such that the functional importance and physiological impact of most tumor genetic alterations remains poorly defined. A major challenge involves the development of analysis methods to uncover biological insights from the data, including the identification of the key mutations that drive cancer and how these events alter cellular regulation.

We have developed Conexic, a novel Bayesian Network-based framework to integrate chromosomal copy number and gene expression data to detect to detect driver genes located in regions that are aberrant in tumors. The underlying assumption is that a driving mutation might be associated with a characteristic gene expression signature representing genes whose expression is modulated by the driver. Thus our score guided approach searches for genes that are both recurrently aberrant and associated with variance of expression patterns across tumor samples. This method not only pinpoints specific regulators within a large aberrant region, but also by associating drivers with gene modules whose expression vary with the driver, provides insight into the physiological roles of drivers and associated genes.

We demonstrated the utility of the CONEXIC framework using a melanoma dataset, our analysis correctly identified known drivers in melanoma (such as MITF) and connected these to many of their known targets, as well as the biological processes they regulate. In addition, it predicted multiple tumor dependencies TBC1D16 and RAB27A in melanoma and showed that tumors highly expressing these genes are dependent on the same gene for growth. Additionally, gene expression in the associated modules is altered following knockdown as predicted by our model. The identity of these drivers suggests that abnormal regulation of protein trafficking is important for cell survival in melanoma and highlights the importance of protein trafficking in this malignancy.

We also present more recent results of applying CONEXIC to additional cancers, including glioblastoma and ovarian cancers, as well as additional phenotypes including invasion and drug resistance. Together, these results demonstrate the ability of integrative Bayesian approaches to identify novel drivers with biological, and possibly therapeutic, importance in cancer.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr SY17-02. doi:10.1158/1538-7445.AM2011-SY17-02

JISTIC: identification of significant targets in cancer

BACKGROUND

Cancer is caused through a multistep process, in which a succession of genetic changes, each conferring a competitive advantage for growth and proliferation, leads to the progressive conversion of normal human cells into malignant cancer cells. Interrogation of cancer genomes holds the promise of understanding this process, thus revolutionizing cancer research and treatment. As datasets measuring copy number aberrations in tumors accumulate, a major challenge has become to distinguish between those mutations that drive the cancer versus those passenger mutations that have no effect.

RESULTS

We present JISTIC, a tool for analyzing datasets of genome-wide copy number variation to identify driver aberrations in cancer. JISTIC is an improvement over the widely used GISTIC algorithm. We compared the performance of JISTIC versus GISTIC on a dataset of glioblastoma copy number variation, JISTIC finds 173 significant regions, whereas GISTIC only finds 103 significant regions. Importantly, the additional regions detected by JISTIC are enriched for oncogenes and genes involved in cell-cycle and proliferation.

CONCLUSIONS

JISTIC is an easy-to-install platform independent implementation of GISTIC that outperforms the original algorithm detecting more relevant candidate genes and regions. The software and documentation are freely available and can be found at: http://www.c2b2.columbia.edu/danapeerlab/html/software.html.

Abstract B70: Conexic: A Bayesian framework to detect drivers and their function uncovers an endosomal signature in melanoma

Genomics is revolutionizing our understanding of cancer biology. Tumor samples assayed for comprehensive chromosomal and gene expression data are accumulating at a staggering rate. A major challenge involves the development of analysis methods to uncover biological insights from these data, including the identification of the key mutations that drive cancer and how these events alter cellular regulation.

We have developed Conexic, a novel computational framework to integrate chromosomal copy number and gene expression data to detect genetic alterations in tumors that drive proliferation, and to model how these alterations perturb normal cell growth/survival. The underlying assumption to our approach is that significantly recurring copy number change, coinciding with its ability to predict the expression patterns varying across tumors, strengthens the evidence of a gene's causative role in cancer. This method not only pinpoints specific regulators within an a large region of copy number variation, but can shed light on the way in which gene regulation is altered

We applyed our Conexic framework to a melanoma dataset (Lin et al, Cancer Research, 2007) comprising 65 paired measurements of gene expression and copy number, with interesting results. Our analysis correctly identified many known ‘driver’ events, while also connecting these to many of their known targets (e.g. MITF). Our global integrative analysis reveals insight into how the drivers alter transcriptional programs. An interesting recurring characteristic is that there are a number of different ways by which drivers can be altered; e.g., the expression of a driver may be altered through copy number variation or other mechanisms, but its influence downstream remains the same.

In addition to confirming the role of known drivers in melanoma, our analysis suggests a number of novel drivers. Most strikingly, these point to a major role in regulation of protein trafficking and endosome biology in this malignancy. These results have linked endosomal processing and sorting to adhesion and survival. Preliminary experimental validation supports three novel drivers including TBC1D16, RAB7A and RAB27A. Together, these results affirm the potential of Conexic to elaborate novel driver modules with biological and possibly therapeutic importance in melanoma and other cancers.

Citation Information: Cancer Res 2009;69(23 Suppl):B70.

Learning signaling network structures with sparsely distributed data

Flow cytometric measurement of signaling protein abundances has proved particularly useful for elucidation of signaling pathway structure. The single cell nature of the data ensures a very large dataset size, providing a statistically robust dataset for structure learning. Moreover, the approach is easily scaled to many conditions in high throughput. However, the technology suffers from a dimensionality constraint: at the cutting edge, only about 12 protein species can be measured per cell, far from sufficient for most signaling pathways. Because the structure learning algorithm (in practice) requires that all variables be measured together simultaneously, this restricts structure learning to the number of variables that constitute the flow cytometer's upper dimensionality limit. To address this problem, we present here an algorithm that enables structure learning for sparsely distributed data, allowing structure learning beyond the measurement technology's upper dimensionality limit for simultaneously measurable variables. The algorithm assesses pairwise (or n-wise) dependencies, constructs "Markov neighborhoods" for each variable based on these dependencies, measures each variable in the context of its neighborhood, and performs structure learning using a constrained search.

Learning a prior on regulatory potential from eQTL data

Genome-wide RNA expression data provide a detailed view of an organism's biological state; hence, a dataset measuring expression variation between genetically diverse individuals (eQTL data) may provide important insights into the genetics of complex traits. However, with data from a relatively small number of individuals, it is difficult to distinguish true causal polymorphisms from the large number of possibilities. The problem is particularly challenging in populations with significant linkage disequilibrium, where traits are often linked to large chromosomal regions containing many genes. Here, we present a novel method, Lirnet, that automatically learns a regulatory potential for each sequence polymorphism, estimating how likely it is to have a significant effect on gene expression. This regulatory potential is defined in terms of “regulatory features”—including the function of the gene and the conservation, type, and position of genetic polymorphisms—that are available for any organism. The extent to which the different features influence the regulatory potential is learned automatically, making Lirnet readily applicable to different datasets, organisms, and feature sets. We apply Lirnet both to the human HapMap eQTL dataset and to a yeast eQTL dataset and provide statistical and biological results demonstrating that Lirnet produces significantly better regulatory programs than other recent approaches. We demonstrate in the yeast data that Lirnet can correctly suggest a specific causal sequence variation within a large, linked chromosomal region. In one example, Lirnet uncovered a novel, experimentally validated connection between Puf3—a sequence-specific RNA binding protein—and P-bodies—cytoplasmic structures that regulate translation and RNA stability—as well as the particular causative polymorphism, a SNP in Mkt1, that induces the variation in the pathway.

Gene expression data of genetically diverse individuals (eQTL data) provide a unique perspective on the effect of genetic variation on cellular pathways. However, the burden of multiple hypotheses, combined with the challenges of linkage disequilibrium, makes it difficult to correctly identify causal polymorphisms. Researchers traditionally apply heuristics for selecting among plausible hypotheses, favoring polymorphisms that are more conserved, that lead to significant amino acid change, or that reside in genes whose function is related to that of the targets. But how do we know how much weight to attribute to different regulatory features? We describe Lirnet, which learns from eQTL data how to weight regulatory features and induce a regulatory potential for sequence variations. Lirnet assesses these weights simultaneously to learning a regulatory network, finding weights that lead to a more predictive network. We show that Lirnet constructs high-accuracy regulatory programs and demonstrate its ability to correctly identify causative polymorphisms. Lirnet can flexibly use any regulatory features, including sequence features that are available for any sequenced organism, and automatically learn their weights in a dataset-specific way. This feature makes it especially advantageous for mammalian systems, where many forms of prior knowledge used in simple model organisms are incomplete or unavailable.

Identifying regulatory mechanisms using individual variation reveals key role for chromatin modification

Sequence polymorphisms affect gene expression by perturbing the complex network of regulatory interactions. We propose a probabilistic method, called Geronemo, which directly aims to identify the mechanism by which genetic changes perturb the regulatory network. Geronemo automatically constructs a set of coregulated genes (modules), whose regulation can involve both sequence variations and expression of regulators. By exploiting the modularity of genetic regulatory systems, Geronemo reveals regulatory relationships that are indiscernible when genes are considered in isolation, allowing the recovery of intricate combinatorial regulation. By incorporating both expression and genotype of regulators, Geronemo captures cases where the effect of sequence variation on its targets is indirect. We applied Geronemo to a data set from the progeny generated by a cross between laboratory BY4716 (BY) and wild RM11-1a (RM) isolates of Saccharomyces cerevisiae. Geronemo produced previously undescribed hypotheses regarding genetic perturbations in the yeast regulatory network, including transcriptional regulation, signal transduction, and chromatin modification. In particular, we find a large number of modules that have both chromosomal characteristics and are regulated by chromatin modification proteins. Indeed, a large fraction of the variance in the expression can be explained by a small number of markers associated with chromatin modifiers. Additional analysis reveals positive selection for sequence evolution of elements in the Swi/Snf chromatin remodeling complex. Overall, our results suggest that a significant part of individual expression variation in yeast arises from evolution of a small number of chromatin structure modifiers.

Causal protein-signaling networks derived from multiparameter single-cell data

Machine learning was applied for the automated derivation of causal influences in cellular signaling networks. This derivation relied on the simultaneous measurement of multiple phosphorylated protein and phospholipid components in thousands of individual primary human immune system cells. Perturbing these cells with molecular interventions drove the ordering of connections between pathway components, wherein Bayesian network computational methods automatically elucidated most of the traditionally reported signaling relationships and predicted novel interpathway network causalities, which we verified experimentally. Reconstruction of network models from physiologically relevant primary single cells might be applied to understanding native-state tissue signaling biology, complex drug actions, and dysfunctional signaling in diseased cells.

Bayesian network analysis of signaling networks: a primer

High-throughput proteomic data can be used to reveal the connectivity of signaling networks and the influences between signaling molecules. We present a primer on the use of Bayesian networks for this task. Bayesian networks have been successfully used to derive causal influences among biological signaling molecules (for example, in the analysis of intracellular multicolor flow cytometry). We discuss ways to automatically derive a Bayesian network model from proteomic data and to interpret the resulting model.

Minreg: inferring an active regulator set

Regulatory relations between genes are an important component of molecular pathways. Here, we devise a novel global method that uses a set of gene expression profiles to find a small set of relevant active regulators, identify the genes that they regulate, and automatically annotate them. We show that our algorithm is capable of handling a large number of genes in a short time and is robust to a wide range of parameters. We apply our method to a combined dataset of S. cerevisiae expression profiles, and validate the resulting model of regulation by cross-validation and extensive biological analysis of the selected regulators and their derived annotations.