Leading Professional Society for Computational Biology and Bioinformatics
Connecting, Training, Empowering, Worldwide

banner

NetBio: Network Biology

COSI Track Presentations

Schedule subject to change
Tuesday, July 23rd
10:20 AM-10:40 AM
Proceedings Presentation: Robust network inference using response logic
  • Torsten Gross, IRI Life Sciences, Humboldt University, Berlin, Germany, Germany
  • Nils Blüthgen, Charité - Universitätsmedizin Berlin, Institut für Pathologie, Berlin, Germany,, Germany

Presentation Overview: Show

Motivation: A major challenge in molecular and cellular biology is to map out the regulatory networks of cells. As regulatory interactions can typically not be directly observed experimentally, various computational methods have been proposed to disentangling direct and indirect effects. Most of these rely on assumptions
that are rarely met or cannot be adapted to a given context.
Results: We present a network inference method that is based on a simple response logic with minimal presumptions. It requires that we can experimentally observe whether or not some of the system’s components respond to perturbations of some other components, and then identifies the directed networks that most accurately account for the observed propagation of the signal. To cope with the intractable number of possible networks, we developed a logic programming approach that can infer networks of hundreds of nodes, while being robust to noisy, heterogeneous or missing data. This allows to directly integrate prior network knowledge and additional constraints such as sparsity. We systematically benchmark our method on KEGG pathways, and show that it outperforms existing approaches in DREAM3 and DREAM4-challenges. Applied to a novel perturbation data set on PI3K and MAPK pathways in isogenic models of a colon cancer cell line, it generates plausible network hypotheses that explain distinct sensitivities towards EGFR inhibitors by different PI3K mutants.
Availability and Implementation: A Python/Answer Set Programming implementation can be accessed at github.com/GrossTor/response-logic. Data and analysis scripts are available at github.com/GrossTor/response-logic-projects.
Contact: nils.bluethgen@charite.de

10:40 AM-11:00 AM
Evolution of resilience in protein interactomes across the tree of life
  • Marinka Zitnik, Stanford University, United States
  • Rok Sosic, Stanford University, United States
  • Marcus Feldman, Stanford University, United States
  • Jure Leskovec, Stanford University, United States

Presentation Overview: Show

Phenotype robustness to environmental fluctuations is a common biological phenomenon. Although most phenotypes involve multiple proteins that interact with each other, the basic principles of how such interactome networks respond to environmental unpredictability and change during evolution are largely unknown. Here we study interactomes of 1,840 species across the tree of life involving a total of 8,762,166 protein-protein interactions. Our study focuses on the resilience of interactomes to network failures and finds that interactomes become more resilient during evolution, meaning that interactomes become more robust to network failures over time. In bacteria, we find that a more resilient interactome is in turn associated with the greater ability of the organism to survive in a more complex, variable, and competitive environment. We find that at the protein family level proteins exhibit a coordinated rewiring of interactions over time and that a resilient interactome arises through the gradual change of the network topology. Our findings have implications for understanding the molecular network structure in the context of both evolution and environment.

11:00 AM-11:20 AM
Coexpression and regulation: the expectation, the observation and the reality
  • Paul Pavlidis, The University of British Columbia, Canada
  • Marjan Farahbod, The University of British Columbia, Canada

Presentation Overview: Show

In transcriptomics, a common assumption is that regulation is the underlying cause of the observed coexpression and that regulatory relationships could be inferred from coexpression links. Here we revisited this assumption by studying differential coexpression between five human tissues to identify the potential cases of regulatory rewiring. We identified many robust tissue-specific links, but found up to 75% of the tissue-specific links to be predictable by the average expression level of the genes. This is contrary to the common belief that much differential coexpression happens in the absence of differential expression. We also found that brain has a particularly high count of tissue-specific links (32% of its total links). Through simulation, we demonstrate that in a heterogeneous bulk tissue, cellular composition variation among the samples could induce variance and coexpression among the genes. To confirm this in the real data, we modelled the variation of genes’ expression levels among the bulk brain samples, using the variation of brain cell-type-marker genes. We show that much of the observed brain specific coexpression is likely to be induced by cellular composition variation among the samples. Our findings raise questions on the potential of the bulk tissue coexpression in predicting regulatory signal.

11:20 AM-11:40 AM
Next-generation biological network alignment
  • Shawn Gu, University of Notre Dame, United States
  • Tijana Milenkovic, University of Notre Dame, United States

Presentation Overview: Show

Genomic sequence alignment has revolutionized understanding of cellular functioning. However, sequence alignment ignores interactions between proteins, which are ultimately what carry out biological processes. Biological network alignment (NA) can fill this gap, transferring functional knowledge between species' conserved molecular network (rather than just sequence) regions. Hence, NA can redefine homology. However, current NA methods do not end up aligning proteins that should be aligned, i.e., that are functionally related. One reason is that traditional NA assumes it is proteins with similar network topologies that are functionally related and should thus be aligned, but we find this not to hold. So, a paradigm shift is needed with how the NA problem is approached. Consequently, we redefine NA as a data-driven framework, which learns from data the relationship between topological relatedness and functional relatedness, without assuming that topological relatedness means topological similarity.
Another possible reason is that traditional NA treats biological data as a homogeneous network. However, biological data are heterogeneous, with different -omics data types capturing different slices of cellular functioning. To handle such data, we generalize homogeneous NA to heterogeneous NA. We find that both data-driven and heterogeneous NA lead to alignments of better functional quality compared to traditional NA.

11:40 AM-12:00 PM
Proceedings Presentation: Inferring signalling dynamics by integrating interventional with observational data
  • Mathias Cardner, ETH Zurich, Switzerland
  • Niko Beerenwinkel, ETH Zurich, Switzerland

Presentation Overview: Show

Motivation: In order to infer a cell signalling network, we generally need interventional data from perturbation experiments. If the perturbation experiments are time-resolved, then signal progression through the network can be inferred. However, such designs are infeasible for large signalling networks, where it is more common to have steady-state perturbation data on the one hand, and a non-interventional time series on the other. Such was the design in a recent experiment investigating the coordination of epithelial–mesenchymal transition (EMT) in murine mammary gland cells. We aimed to infer the underlying signalling network of transcription factors and microRNAs coordinating EMT, as well as the signal progression during EMT.

Results: In the context of nested effects models, we developed a method for integrating perturbation data with a non-interventional time series. We applied the model to RNA sequencing data obtained from an EMT experiment. Part of the network inferred from RNA interference was validated experimentally using luciferase reporter assays. Our model extension is formulated as an integer linear programme, which can be solved efficiently using heuristic algorithms. This extension allowed us to infer the signal progression through the network during an EMT time course, and thereby assess when each regulator is necessary for EMT to advance.

12:00 PM-12:40 PM
Protein-protein association networks and their use in complementing functional pathway enrichment analysis
  • Christian von Mering

Presentation Overview: Show

For many high-throughput omics studies, statistical enrichment analysis is an essential step towards highlighting functional processes or pathways that are responsible for a given phenotype or mechanism of interest. However, the currently annotated pathway knowledge is still incomplete, and furthermore it is not always clear how the various interconnected pathways should be delineated from each other. Protein-protein association networks effectively describe a superset of pathways and pathway knowledge, but are more error-prone and much less structured. Here, we use the STRING database of functional associations to complement traditional enrichment analysis. The STRING network is hierarchically pre-clustered on a diffusion distance metric, and the resulting partitioning is used as an additional set of "pathways" for the functional enrichment analysis. I describe how this relates to traditional pathway annotations, and to what extent it enables the discovery of novel subclusters for phenotypes of interest.

2:00 PM-2:20 PM
PIMKL: Pathway Induced Multiple Kernel Learning
  • Matteo Manica, IBM, Switzerland
  • Joris Cadow, IBM, Switzerland
  • Roland Mathis, Telepathy Labs, Switzerland
  • María Rodríguez Martínez, IBM, Switzerland

Presentation Overview: Show

Reliable identification of molecular biomarkers is essential for accurate patient stratification. While state-of-the-art machine learning approaches for sample classification continue to push boundaries in terms of performance, most of these methods are not able to integrate different data types and lack generalization power, limiting their application in a clinical setting.
Furthermore, many methods behave as black boxes, and we have very little understanding about the mechanisms that lead to the prediction. In health care however, providing explanations about the molecular factors and phenotypes that are driving the classification is crucial to build trust in the performance of the predictive system.
We propose Pathway Induced Multiple Kernel Learning (PIMKL), a methodology to reliably classify samples that can also help gain insights into the molecular mechanisms that underlie the classification. PIMKL exploits prior knowledge in the form of a molecular interaction network and annotated gene sets, by optimizing a mixture of pathway-induced kernels using a Multiple Kernel Learning (MKL) algorithm. After optimizing the combination of kernels to predict a specific phenotype, the model provides a stable molecular signature that can be interpreted in the light of the ingested prior knowledge and that can be used in transfer learning tasks.

2:20 PM-2:40 PM
Towards a data-integrated cell
  • Julia Petschnigg, University College London, United Kingdom
  • Sam Windels, University College London, United Kingdom
  • Janez Povh, University of Ljubljana, Slovenia
  • Harry Hemmingway, University College London, United Kingdom
  • Robin Ketteler, University College London, United Kingdom
  • Natasa Przulj, Barcelona Supercomputing Center (BSC), Spain
  • Noel Malod-Dognin, Barcelona Supercomputing Center, Spain

Presentation Overview: Show

We are flooded with large scale molecular data capturing complementary aspects of the functioning of a cell. To enable new discoveries, we propose a novel, data-driven concept of an integrated cell, iCell. Also, we introduce a computational prototype of an iCell, which integrates three omics, tissue-specific molecular interaction network types: protein-protein interactions, gene co-expressions, and genetic interactions. We apply our framework and construct iCells of four cancers, breast, prostate, lung, and colorectal, as well as of the corresponding tissue controls. Comparison between cancer and control iCells allows us to uncover the most rewired genes in cancer that do not appear as different in any of the constituent data types. Many of these genes are of unknown function. We biologically validate that they have a role in cancer by knockdown experiments followed by cell viability assays. We find additional support through Kaplan-Meier survival curves of thousands of patients. Finally, we extend this analysis to twenty different cancer types to uncover new pan-cancer genes. Our methodology is universal and enables integrative omics comparisons of diverse data over cells and tissues.

2:40 PM-3:00 PM
Integrative network-based approach identifies gene communities in COPD
  • Roberta Marino, Department of Clinical and Biological Sciences, University of Turin, Italy, Italy
  • Yousef El Aalamat, GSK, Rixensart, Belgium, Belgium
  • Vanesa Bol, GSK, Rixensart, Belgium, Belgium
  • Duccio Medini, GSK, Siena, Italy, Italy
  • Christophe Lambert, GSK, Rixensart, Belgium, Belgium
  • Alessandro Muzzi, GSK, Siena, Italy, Italy
  • Michele Caselle, Torino university, Italy

Presentation Overview: Show

Chronic Obstructive Pulmonary Disease (COPD) is characterised by exacerbation phases alternating to stable conditions. To identify gene communities modulated in these particular phases, we applied a multi-network strategy, integrating a gene expression network with different layers of omics information. We implemented the pipeline for blood gene expression data from the AERIS clinical study (NCT01360398), which has observed 127 COPD patients for two years. For each explored condition (exacerbation and stable state) a gene co-expression network was built and integrated with a co-regulation network, a human protein-protein interaction network, a transcription factor and a microRNA co-targeting network. We applied Infomap algorithm to detect communities. Because of the stochasticity of the algorithm we introduced a robustness step assessment, by repeating multiple times the community detection to compute a final consensus clustering and therefore determine presence of relevant and interacting communities of genes. Using our pipeline, we identified co-regulated genes in blood samples taken at regular and at exacerbation visits. Specific metabolic functional responses are remarkably enriched in both conditions while co-regulation of a limited set of functions and potential targets of miRNA were enriched/characteristic of exacerbation phases. These last ones are not detected by the classical GSEA method.

Funding: GlaxoSmithKline Biologicals SA

3:00 PM-3:20 PM
Single-sample network modeling identifies regulation of PD1 signaling associated with glioblastoma survival
  • John Quackenbush, Harvard Chan School of Public Health, United States
  • Marieke Kuijjer, Centre for Molecular Medicine Norway, University of Oslo, Norway
  • Camila Lopes-Ramos, Harvard Chan School of Public Health, United States
  • Tess Brunner, Wesleyan University, United States

Presentation Overview: Show

Glioblastoma is an aggressive cancer of the brain and spine. While analysis of glioblastoma `omics data has somewhat improved our understanding of the disease, it has not led to direct improvement of patient survival. Cancer survival is often characterized by differences in expression of particular genes, but the mechanisms that drive these differences are generally unknown. We set out to model the regulatory mechanisms that associate with glioblastoma survival. We used our previously developed methods PANDA and LIONESS to model individual patient (n=525 and 431) gene regulatory networks using data from two different expression platforms from The Cancer Genome Atlas. We performed a comparative network analysis between patients with long- and short-term survival by using a LIMMA analysis on gene degree, correcting for patient age, sex, and neoadjuvant treatment status. We used Gene Set Enrichment Analysis to identify 7 pathways associated with survival, all of which were involved in immune system signaling. We validated one pathway, PD1 signaling, in an independent dataset from the German Glioma Society (n=70). Transcriptional repression of genes in this pathway—for which treatment options are available—was lost in short-term survivors. These results underscore the importance of analyzing gene regulatory networks in cancer.

3:20 PM-3:30 PM
scNetViz: A Cytoscape App for the Network Analysis and Visualization of scRNA-Seq Data
  • Alexander Pico, Gladstone Institutes, United States
  • Elaine Meng, UCSF RBVI, United States
  • Javier Diaz-Mejia, Princess Margaret Cancer Centre, Canada
  • Gary Bader, University of Toronto, Canada
  • John H. Morris, University of California, San Francisco, United States

Presentation Overview: Show

Recent advances in scRNA-seq technology have led to significant efforts to store and share the resulting data. For example, the EBI has extended Expression Atlas to include single-cell data, and the Chan-Zuckerberg Initiative has launched a major project to build a Human Cell Atlas, which will use single-cell transcriptomics along with images and other data to categorize and identify all of the cell types in the human body. Systems such as Seurat and scanpy offer pipelines for the analysis of scRNA-seq data, but few tools exist for its integration with other -omics data or for its downstream analysis within the context of protein networks or pathways. The construction of network models from single-cell data is a crucial step in exploratory analysis and interpretation, offering a host of graph-based analytical and visualization opportunities. Here we present scNetViz, a Cytoscape App for the network analysis and visualization of single-cell RNA-seq data.

3:30 PM-3:40 PM
Omics Visualizer: a Cytoscape App to visualize omics data
  • Marc Legeay, Novo Nordisk Foundation Center for Protein Research, Denmark
  • Nadezhda T. Doncheva, University of Copenhagen, Denmark
  • John H. Morris, University of California, San Francisco, United States
  • Lars J. Jensen, University of Copenhagen, Denmark

Presentation Overview: Show

Cytoscape is an open-source software used to analyze and visualize networks. In addition to being able to import networks from a variety of sources, Cytoscape allows users to import tabular node data and visualize it onto networks. Unfortunately, such data tables can only contain one row of data per node, whereas omics data often have multiple rows for the same gene or protein, representing different post-translational modification sites, peptides, splice isoforms, or conditions. However, Cytoscape has an API that allows developers to make apps that extend its functionality. Here, we present a new app, Omics Visualizer, that allows users to import data tables with several rows referring to the same node, connect them to one or more networks, and visualize the connected data onto networks. Omics Visualizer uses the Cytoscape enhancedGraphics app to draw charts in the nodes (pie charts) or around the nodes (donut charts). If the user does not provide a network, Omics Visualizer can retrieve one from the STRING database using the Cytoscape stringApp. The app is freely available at http://apps.cytoscape.org/apps/omicsvisualizer.

3:40 PM-3:50 PM
Community challenge assesses network module identification methods across complex diseases
  • Daniel Marbach, Roche Innovation Center Basel, Switzerland
  • Ted Natoli, Harvard University, United States
  • Rajiv Narayan, Harvard University, United States
  • Aravind Subramanian, Harvard University, United States
  • Gustavo Stolovitzky, IBM, United States
  • Zoltán Kutalik, Lausanne University Hospital, Switzerland
  • Lenore J. Cowen, Tufts University, United States
  • Sven Bergmann, University of Lausanne, Switzerland
  • Julio Saez-Rodriguez, Institute of Computational Biomedicine, Heidelberg University, Germany
  • Johnathan Mercer, Harvard University, United States
  • Xiaozhe Hu, Tufts University, United States
  • Sarvenaz Choobdar, University of Lausanne, Switzerland
  • Mehmet E. Ahsen, Icahn School of Medicine at Mount Sinai, United States
  • Jake Crawford, Tufts University, United States
  • Mattia Tomasoni, University of Lausanne, Switzerland
  • Tao Fang, Roche Innovation Center Basel, Switzerland
  • David Lamparter, University of Lausanne, Switzerland
  • Junyuan Lin, Tufts University, United States
  • Benjamin Hescott, Northeastern University, United States
  • Jitao David Zhang, Roche Innovation Center Basel, F. Hoffmann-La-Roche AG,, Switzerland

Presentation Overview: Show

Many bioinformatics methods have been proposed for reducing the complexity of large gene or protein networks into relevant subnetworks or modules. Yet, how such methods compare to each other in terms of their ability to identify disease-relevant modules in different types of networks remains poorly understood. We launched the “Disease Module Identification DREAM Challenge”, an open competition to comprehensively assess module identification methods across diverse protein-protein interaction, signaling, gene co-expression, homology, and cancer-gene networks. Predicted network modules were tested for association with complex traits and diseases using a unique collection of 180 genome-wide association studies (GWAS). Our critical assessment of 75 contributed module identification methods reveals novel top-performing algorithms, which recover complementary trait-associated modules. We find that most of these modules correspond to core disease-relevant pathways, which often comprise therapeutic targets. This community challenge establishes benchmarks, tools and guidelines for molecular network analysis to study human disease biology (https://synapse.org/modulechallenge).

3:50 PM-4:00 PM
GUILDify v2.0: A tool to identify molecular networks underlying human diseases, their comorbidities and their druggable targets
  • Ferran Sanz, GRIB (IMIM-UPF), Spain
  • Narcis Fernandez-Fuentes, Aberystwyth University, United Kingdom
  • Emre Guney, GRIB (IMIM-UPF), Spain
  • Joaquim Aguirre-Plans, GRIB (IMIM-UPF), Spain
  • Janet Piñero, GRIB (IMIM-UPF), Spain
  • Laura I. Furlong, GRIB (IMIM-UPF), Spain
  • Baldo Oliva, GRIB (IMIM-UPF), Spain

Presentation Overview: Show

The genetic basis of complex diseases involves alterations on multiple genes. Unravelling the interplay between these genetic factors is key to the discovery of new biomarkers and treatments. In 2014, we introduced GUILDify, a web server that searches for genes associated to diseases, finds novel disease-genes applying various network-based prioritisation algorithms and proposes candidate drugs. Here, we present GUILDify v2.0, a major update and improvement of the original method, where we have included protein interaction data for seven species and 22 human tissues and incorporated the disease-gene associations from DisGeNET. To infer potential disease relationships associated with multi-morbidities, we introduced a novel feature for estimating the genetic and functional overlap of two diseases using the top-ranking genes and the associated enrichment of biological functions and pathways (as defined by GO and Reactome). The analysis of this overlap helps to identify the mechanistic role of genes and protein-protein interactions in comorbidities. Finally, we provided an R package, guildifyR, to facilitate programmatic access to GUILDify v2.0 (http://sbi.upf.edu/guildify2). The research has been recently published in Journal of Molecular Biology (doi: 10.1016/j.jmb.2019.02.027).

4:40 PM-5:00 PM
DiseaseScope: Automatic Construction and Interpretation of Hierarchical Disease Models
  • Samson Fong, University of California San Diego, United States
  • Sui Huang, Institute for Systems Biology, United States
  • Dexter Pratt, University of California San Diego, United States
  • Keiichiro Ono, University of California San Diego, United States
  • Christopher Churas, University of California San Diego, United States
  • Michael Yu, University of Chicago, United States
  • Sheng Wang, Stanford University, United States
  • Fan Zheng, University of California San Diego, United States
  • Jianzhu Ma, University of California San Diego, United States
  • Theo Knijnenburg, Institute for Systems Biology, United States
  • Trey Ideker, Department of Medicine, University of California, San Diego, United States

Presentation Overview: Show

Determining the genetic basis of diseases is critical to developing therapeutic strategies. We have previously developed approaches to integrate rich and diverse sets of omics data into interpretable, hierarchical models and have found that they can both recapitulate known cellular subsystems and guide discovery of new ones. To enable systematic discovery of biomedical knowledge, we have built DiseaseScope, a service that automatically organizes high-throughput gene-gene interaction data into interactive hierarchical models. This method takes a disease name and returns biological information at multiple scales including a core set of disease-associated genes, interactions that form disease-relevant pathways and the hierarchical organization of these pathways. Furthermore, to elucidate how each gene cluster is related to the disease, DiseaseScope includes two interpretation tools: HiView Lens to explore the underlying structure of individual modules by overlaying additional networks and NetAnt to determine what biomedical concepts connect the gene module to disease by proposing mechanistic pathways to pathogenesis. Although the pipeline is automatic, each module is a self-contained service that can be invoked independently, allowing users to form custom applications. Together, DiseaseScope aggregates across massive amounts of biological knowledge about diseases and organize the knowledge to guide discovery.

5:00 PM-5:20 PM
Identifying Drug Sensitivity Subnetworks with NETPHLEX
  • Damian Wójtowicz, National Institutes of Health, NCBI, United States
  • Teresa Przytycka, National Center of Biotechnology Information, NLM, NIH, United States
  • Yoo-Ah Kim, National Institue of Health, United States
  • Rebecca Sarto Basso, University of California, Berkeley, United States
  • Dorit S. Hochbaum, Department of Industrial Engineering and Operations Research, University of California at Berkeley, CA, USA, United States
  • Fabio Vandin, University of Padova, Italy

Presentation Overview: Show

Phenotypic heterogeneity in cancer is often caused by different genetic alterations. Understanding such phenotype-genotype relationships is fundamental for the advance of personalized medicines. Phenotype-genotype relationships in cancer can be better interpreted in a pathway-centric view, in which genetic alterations in the disease are considered from the context of dysregulated pathways. However, most of pathway identification methods in cancer focus on finding subnetworks that include general cancer drivers or are associated with discrete features, hence cannot be directly applied for the analysis of continuous features such as drug response. On the other hand, existing genome wide association approaches do not fully utilize the complex proprieties of cancer mutations. To address these challenges, we propose NETPHLEX (NETwork-to-PHenotype mapping Leveraging EXclusivity), which aims to identify subnetworks of mutated genes that are collectively associated with continuous cancer phenotypes. We formulate the problem as an integer linear program and solve it optimally to obtain a connected set of mutated genes maximizing the association. Analyzing a cell line drug response dataset, we identified sensitivity associated subnetworks for a large set of drugs. NETPHLEX can be used to identify subnetworks associated with any continuous phenotypes beyond drug response data.

5:20 PM-6:00 PM
Connecting genomics and network properties of genes relevant for disease and drug response
  • Laura I. Furlong, GRIB (IMIM-UPF), Spain

Presentation Overview: Show

Network analysis has been used to understand the relationship between genotype and phenotype. In the context of disease genomics, characterizing the topological role of disease genes in biological networks has the potential to shed light on disease mechanisms and to reveal both new candidate disease genes and therapeutic targets. Previous studies addressing the network properties of disease genes have produced contradictory results. We propose that these discrepancies can be addressed by a careful selection of different classes of disease genes and by using a multi-scale approach for network analysis. Moreover, we describe a relationship between the role that different classes of disease genes play in cellular networks and their tolerance to deleterious genomic variation. We use a similar approach to characterize genes relevant to drug response, in order to shed light on the mode of action of drugs and their toxicity. We show how different classes of proteins involved in the therapeutic effect of drugs and in their adverse effects differ on their transcriptomic, genomic and multi-scale network patterns. These findings highlight distinctive properties of proteins related to drug action, which could be applied to prioritize drugs with fewer probabilities of causing side effects.