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

banner

NetBio COSI Track Presentations

Attention Conference Presenters - please review the Speaker Information Page available here
NetBio Welcome
Date: Sunday, July 23
Time: 10:00 a.m. - 10:15 a.m.
Room: Panorama
  • Alexander Pico, Gladstone Institutes, United States

Presentation Overview: Show

Welcome message and overview of the day's agenda.

Alignment of dynamic networks
Date: Sunday, July 23
Time: 10:15 a.m. - 10:30 a.m.
Room: Panorama
  • Tijana Milenkovic, University of Notre Dame, United States
  • Dominic Critchlow, University of Notre Dame, United States
  • Vipin Vijayan, University of Notre Dame, United States

Presentation Overview: Show

Networks can model real-world systems in a variety of domains. Network alignment (NA) aims to find a node mapping that conserves similar regions between compared networks. NA is applicable to many fields, including computational biology, where NA can guide the transfer of biological knowledge from well- to poorly-studied species across aligned network regions. Existing NA methods can only align static networks. However, most complex real-world systems evolve over time and should thus be modeled as dynamic networks. We hypothesize that aligning dynamic network representations of evolving systems will produce superior alignments compared to aligning the systems' static network representations, as is currently done. For this purpose, we introduce the first ever dynamic NA method, DynaMAGNA++. This proof-of-concept dynamic NA method is an extension of a state-of-the-art static NA method, MAGNA++. Even though both MAGNA++ and DynaMAGNA++ optimize edge as well as node conservation across the aligned networks, MAGNA++ conserves static edges and similarity between static node neighborhoods, while DynaMAGNA++ conserves dynamic edges (events) and similarity between evolving node neighborhoods. For this purpose, we introduce the first ever measure of dynamic edge conservation and rely on our recent measure of dynamic node conservation. Importantly, the two dynamic conservation measures can be optimized using any state-of-the-art NA method and not just MAGNA++. We confirm our hypothesis that dynamic NA is superior to static NA, under fair comparison conditions, on synthetic and real-world networks, in computational biology and social network domains. DynaMAGNA++ is parallelized and it includes a user-friendly graphical interface.

Image-based Spatiotemporal Causality Inference for Protein Signaling Networks
Date: Sunday, July 23
Time: 10:30 a.m. - 10:45 a.m.
Room: Panorama
  • Christoph Wülfing, University of Bristol, United Kingdom
  • Xiongtao Ruan, Carnegie Mellon University, United States
  • Robert F. Murphy, Carnegie Mellon University, United States

Presentation Overview: Show

Motivation: Efforts to model how signaling and regulatory networks work in cells have largely either not considered spatial organization or have used compartmental models with minimal spatial resolution. Fluorescence microscopy provides the ability to monitor the spatiotemporal distribution of many molecules during signaling events, but as of yet no methods have been described for large scale image analysis to learn a complex protein regulatory network. Here we present and evaluate methods for identifying how changes in concentration in one cell region influence concentration of other proteins in other regions.

Results: Using 3D confocal microscope movies of GFP-tagged T cells undergoing costimulation, we learned models containing putative causal relationships among 12 proteins involved in T cell signaling. The models included both relationships consistent with current knowledge and novel predictions deserving further exploration. Further, when these models were applied to the initial frames of movies of T cells that had been only partially stimulated, they predicted the localization of a number of proteins at later times with statistically significant accuracy. The methods, consisting of spatiotemporal alignment, automated region identification, and causal inference, are anticipated to be applicable to a number of biological systems.

Predicting multicellular function through multi-layer tissue networks
Date: Sunday, July 23
Time: 10:45 a.m. - 11:00 a.m.
Room: Panorama
  • Jure Leskovec, Stanford University, United States
  • Marinka Zitnik, Stanford University, United States

Presentation Overview: Show

Motivation: Understanding functions of proteins in specific human tissues is essential for insights into disease diagnostics and therapeutics, yet prediction of tissue-specific cellular function remains a critical challenge for biomedicine.

Results: Here we present OhmNet, a hierarchy-aware unsupervised node feature learning approach for multi-layer tissue networks. We build a multi-layer network, where each layer represents molecular interactions in a different human tissue. OhmNet then automatically learns a mapping of proteins, represented as nodes, to a neural embedding based low-dimensional space of features. OhmNet encourages sharing of similar features among proteins with similar network neighborhoods and among proteins activated in similar tissues. The algorithm generalizes prior work, which generally ignores relationships between tissues, by modeling tissue organization with a rich multiscale hierarchy. We use OhmNet to study multicellular function in a multi-layer protein interaction network of 107 human tissues. In 48 tissues with known tissue-specific cellular functions, OhmNet provides more accurate predictions of cellular function than alternative approaches, and also generates more accurate hypotheses about tissue-specific protein actions. We show that taking into account the tissue hierarchy leads to improved predictive power. Remarkably, we also demonstrate that it is possible to leverage the tissue hierarchy in order to effectively transfer cellular functions to a functionally uncharacterized tissue. Overall, OhmNet moves from flat networks to multiscale models able to predict a range of phenotypes spanning cellular subsystems.

Incorporating Interaction Networks into the Determination of Functionally Related Hit Genes in Genomic Experiments with Markov Random Fields
Date: Sunday, July 23
Time: 11:00 a.m. - 11:15 a.m.
Room: Panorama
  • Laurent Guyon, CEA, France
  • J. Pablo Radicella, CEA, France
  • Anna Campalans, CEA, France
  • Guillaume Pinna, CEA, France
  • Jaakko Nevalainen, University of Turku, Finland
  • Sean Robinson, University of Turku, Finland

Presentation Overview: Show

Motivation: Incorporating gene interaction data into the identification of ‘hit’ genes in genomic experiments is a well-established approach leveraging the ‘guilt by association’ assumption to obtain a network based hit list of functionally related genes. We aim to develop a method to allow for multivariate gene scores and multiple hit labels in order to extend the analysis of genomic screening data within such an approach.

Results: We propose a Markov random field based method to achieve our aim and show that the particular advantages of our method compared to those currently used lead to new insights in previously analysed data as well as for our own motivating data. Our method additionally achieves the best performance in an independent simulation experiment. The real data applications we consider comprise of a survival analysis and differential expression experiment and a cell-based RNA interference functional screen.

Availability: We provide all of the data and code related to the results in the paper.

Active Interaction Mapping reveals the hierarchical organization of autophagy
Date: Sunday, July 23
Time: 11:15 a.m. - 11:30 a.m.
Room: Panorama
  • Trey Ideker, University of California, San Diego, United States
  • J. Michael Cherry, Stanford University, United States
  • Suresh Subramani, University of California, San Diego, United States
  • Rama Balakrishnan, Stanford University, United States
  • Mitchell Flagg, University of California, San Diego, United States
  • Katherine Licon, University of California, San Diego, United States
  • Barry Demchak, University of California, San Diego, United States
  • Keiichiro Ono, University of California, San Diego, United States
  • Michael Ku Yu, University of California, San Diego, United States
  • Koyel Mitra, University of California, San Diego, United States
  • Jean-Claude Farre, University of California, San Diego, United States
  • Michael Kramer, University of California, San Diego, United States

Presentation Overview: Show

We have developed a general progressive procedure, Active Interaction Mapping, to guide assembly of the hierarchy of functions encoding any biological system. Using this process, we assemble an ontology of functions comprising autophagy, a central recycling process implicated in numerous diseases. A first-generation model, built from existing gene networks in Saccharomyces, captures most known autophagy components in broad relation to vesicle transport, cell cycle and stress response. Systematic analysis identifies synthetic-lethal interactions as most informative for further experiments; consequently, we saturate the model with 156,364 such measurements across autophagy-activating conditions. These targeted interactions provide more information about autophagy than all previous datasets, producing a second-generation ontology of 220 functions. Approximately half are new, in which we confirm roles for Gyp1 at the phagophore-assembly site, Atg24 in cargo engulfment, Atg26 in cytoplasm- to-vacuole targeting, and Ssd1, Did4 and others in selective and non-selective autophagy. The procedure and autophagy hierarchy are at http://atgo.ucsd.edu/.

PathSys: Integrating pathway curation, profiling methods, and public repositories: An infrastructure for functional molecular data sharing
Date: Sunday, July 23
Time: 11:30 a.m. - 11:45 a.m.
Room: Panorama
  • Sokratis Kariotis, University of Sheffield, United Kingdom

Presentation Overview: Show

Data integration at the level of high dimensional molecular interrogation is confounded by the diaspora of platforms and annotations of molecular events. To unify interpretation of functional activity within and between samples, we are developing a suite of tools that confer a highly standardised representation of pathway activity, networked pathway activity correlation, and pathway/disease/drug interaction. We have discovered that using the concept of higher order gene set interactions, using gene sets as the unit of comparison we are able to unify very large sets of data without a reliance on geneset overlap. Pathprint is the most developed of our set of tools: a functional approach that compares gene expression as a tertiary summary statistic for each canonical pathway, generating a set of pathway activities, networks and transcriptionally regulated targets. It compares a sample against a background of thousands of arrays to yield a relative activity for each pathway tested. It can be applied universally to gene expression profiles across species. Integration of large-scale profiling methods and curation of the public repository overcomes platform, species and batch effects to yield a standard measure of functional distance between experiments. Pathprint version (v2.0), shortly available through Bioconductor, includes 35 platforms, with new additions effectively increasing the number of covered arrays to 446,708; providing a 4x increase in background for pathway comparisons. Pathprint is utilised by the Harvard Stem Cell commons (http://stemcellcommons.org) as part of standardisation for representation and comparisons of stem cell systems. It is being implemented within the Genometranslationcommons (https://beta.genometranslationcommons.org//#/) at the University of Sheffield and the CureADCircuitscommons (in dev) as part of a Harvard/MIT/Sheffield consortium investigating regulation of genes associated with Alzheimer’s. PCxN (namely the Pathway Co-Expression Network) (Hide, Winston (2015): PCxN the Pathway co-activity Map. figshare. https://doi.org/10.6084/m9.figshare.1589792.v4) is an online web resource which allows the discovery of correlation relationships between groups of pathways or gene sets drawn from the MsigDB and Pathprint collections. The tool provides users the ability to explore a static extendable network by focusing on single pathways and their most correlated neighbours, as well as identifying relationships between groups of pathways shown to be enriched in the collections by gene set enrichment. Analyses can be viewed and exported through a heatmap, a correlation network and gene/network tables. PCxN is employed as part of the CureADCircuits consortium (publication in prep) and is deployed for interpretation of network and pathway relationships by the AMP-AD consortium. PDN (Pathway Drug Network), currently in development, relies on a network, made up of the expression correlation between each of 16,150 drug, disease and pathway gene signatures across 58,475 publicly available human microarrays (Affymetrix HGU133 Plus2) collected from the Comparative Toxicogenomics Database, PharmGKB, GeneSigDB, Wikipathways, KEGG, Netpath, Reactome, and Connectivity Map. PDN aims to utilize pathway – drug relationships to identify drug leads and to prioritise pathways that can be targeted in relationships to disease profiles. Its prototype has been successfully used together with Pathprint at Harvard School of Public Health in (Joachim R., Altschuler G., Hutchinson J., Wong H., Hide W., Kobzik L.: Pathway- centered Analysis of the Relative Resistance of Children to Sepsis Mortality, in preparation). We have shown that PDN has a substantially higher rate of positives (p<0.01) when compared to a purely gene-level ConnectivityMap analysis (54% vs. 27%). In direct testing of drug candidates using an endotoxemia model of murine sepsis, 5 of 10 compounds improved survival. Taken as a whole, these approaches provide the first standardised approach to representation of systems biology with significant new insight into the systems level interpretation of gene set activity and correlation between genesets.

A novel method for network crosstalk analysis that improves accuracy of pathway annotation
Date: Sunday, July 23
Time: 11:45 a.m. - 11:50 a.m.
Room: Panorama
  • Dimitri Guala, Stockholm University, Sweden
  • Christoph Ogris, Stockholm University, Sweden
  • Erik Sonnhammer, Stockholm University, Sweden

Presentation Overview: Show

A plethora of tools exist to identify significant enrichment of pathways for a set of differentially expressed genes. Most tools analyze gene overlap between gene sets and are therefore severely hampered by the current state of pathway annotation, yet at the same time they run a high risk of false assignments. A way to improve both true positive and false positive rates is to use a functional association network and instead look for enrichment of network connections between gene sets. We present a new network crosstalk analysis method BinoX that determines the statistical significance of network link enrichment or depletion between gene sets, using the binomial distribution. This is a much more appropriate statistical model than previous methods have employed, and as a result BinoX yields substantially better true positive and false positive rates than was possible before. A number of benchmarks were performed to assess the accuracy of BinoX and competing methods. We will demonstrate examples of how BinoX finds many biologically meaningful pathway annotations for gene sets from cancer and other diseases, which are not found by other methods. BinoX is available at http://sonnhammer.org/BinoX. For online pathway analysis of single gene sets, we have set up a web server http://PathwAX.sbc.su.se which applies the BinoX algorithm to KEGG pathways and FunCoup networks. The system was optimized for speed and allows interactive web usage.

Drug Response Prediction as a Link Prediction Problem
Date: Sunday, July 23
Time: 11:50 a.m. - 11:55 a.m.
Room: Panorama
  • Mehmet Koyuturk, Case Western Reserve University, United States
  • Mustafa Coskun, Case Western Reserve University, United States
  • Zachary Stanfield, Case Western Reserve University, United States

Presentation Overview: Show

Drug response prediction is a well-studied problem in which the molecular profile of a given sample is used to predict the effect of a given drug on that sample. Effective solutions to this problem hold the key for precision medicine. In cancer research, genomic data from cell lines are often utilized as features to develop machine learning models predictive of drug response. Today, multiple public resources provide drug response data on large cohorts of cell lines. These resources include the Genomics of Drug Sensitivity in Cancer (GDSC) and the Cancer Cell Line Encyclopedia (CCLE). Molecular network data is frequently utilized for various predictive tasks that involve using "omic" data as features. Networks provide a functional context for the integration of molecular features, thereby resulting in robust and reproducible predictive models. However, inclusion of network data increases dimensionality and poses additional challenges for such machine learning tasks as feature construction, feature selection, and model validation. To overcome these challenges, we here formulate drug response prediction as a link prediction problem. This approach results in a simple formulation through representation of multiple data types into a heterogeneous network that integrates the relationships between drugs, cell lines, molecular aberrations, and functional associations among biomolecules. We compute “network profiles" for drugs and cell lines, which represent their location in this network. The associations between these profiles are then used to predict links between drugs and cell lines. Through leave-one-out cross validation on the GDSC data set, we show that this approach leads to the classification of sensitive and resistant cell line-drug pairs with 88% accuracy. We also perform cross- classification studies by using profiles derived from the GDSC network to predict links in the CCLE network, and observe 85% prediction accuracy across different datasets. We show that our method outperforms a state-of-the-art machine learning algorithm for the prediction tasks of classifying drugs for individual cell lines and cell lines for individual drugs. Finally, the calculated network profiles are examined to determine their biological relevance, and we find functional terms relating to a drug's target are seen to be differentially enriched between the resistant and sensitive profiles.

NetProphet 2.0: Mapping Transcription Factor Networks by Exploiting Scalable Data Resources
Date: Sunday, July 23
Time: 11:55 a.m. - 12:00 p.m.
Room: Panorama
  • Hien-Haw Liow, Washington University, United States
  • Michael R. Brent, wustl, United States
  • Ezekiel Maier, Washington University, United States
  • Yiming Kang, Washington University, United States

Presentation Overview: Show

Motivation: Cells process information, in part, through transcription factor (TF) networks, which control the rates at which individual genes produce their products. A TF network map is a graph that indicates which TFs bind and directly regulate each gene. Previous work has described both network mapping algorithms that rely exclusively on gene expression data and “integrative” algorithms that exploit a wide range of data sources including chromatin immunoprecipitation sequencing (ChIP-seq) of many TFs, genome-wide chromatin marks, and binding specificities for many TFs determined in vitro. However, such resources are available only for a few major model systems and cannot be easily replicated for new organisms or cell types. Results: We present NetProphet 2.0, a “data light” algorithm for TF network mapping (Fig. 1A) and show that it is more accurate at identifying direct targets of TFs than other, similarly data light algorithms (Fig. 1B). It work by exploiting three principles. First, combining multiple approaches to network mapping from expression data can improve accuracy relative to the constituent approaches. Second, TFs with similar DNA binding domains bind similar sets of target genes. Third, even a noisy, preliminary network map can be used to infer DNA binding specificities from promoter sequences and these inferred specificities can be used to further improve the accuracy of the network map. The result is the only algorithm that has been shown to be accurate on animal networks. Approach: NetProphet 2.0 comprises 6 modules (Fig. 1A), 5 of which take advantage of information derived from gene expression profiling or genome sequencing. The output of each module is a score matrix (TFs x genes). The score vector (row) for a TF represents the strength of evidence that the TF regulates each potential target gene. A discrete graph structure can be constructed by thresholding edge scores. Module A (Fig. 1Aa) is NetProphet 1.0, as previously described. Module B (Fig. 1Ab) constructs an independent network from the gene expression data by using Bayesian Additive Regression Trees (BART), a machine learning algorithm. For each gene, Module B trains a separate BART model to predict its RNA level as a function of the RNA levels of all TFs. It then simulates the effect on the predicted RNA level when each TF’s RNA level is varied between its min and max observed levels. The difference between these predicted gene expression levels is the score of the TF-target pair. Module C (Fig. 1Ac) capitalizes on the fact that TFs with similar DNA binding domains (DBDs) tend to bind similar sets of target genes. It replaces the score matrix row for each TF by a weighted average of rows for other TFs with similar DBDs. Each row is weighted according to how similar the DBD of its TF is to the DBD of the row being replaced. The predicted amino acid sequence of the DBD can be obtained from automated genome annotation. The outputs of modules A and B are independently passed through Module C, then combined into a single score matrix by Module D (Fig. 1Ad), which uses quantile normalization to match their score distributions. Module E infers the DNA-binding motif of each TF as the motif whose presence in a promoter best distinguishes high scoring (likely) target genes from low scoring (unlikely) target genes. Module F scans the inferred motifs over the promoters of all genes and computes a score reflecting the strength of evidence that the TF binds the promoter. The resulting score matrix is then combined with the input score matrix by using module D again. In a final step, the combined matrix is passed through module C again. Availability: https://github.com/yiming-kang/NetProphet_2.0.

NetBio Poster Highlights
Date: Sunday, July 23
Time: 12:00 p.m. - 12:30 p.m.
Room: Panorama
  • Poster Authors

Presentation Overview: Show

One poster, one slide, one minute each.

Lunch
Date: Sunday, July 23
Time: 12:30 p.m. - 2:00 p.m.
Room: Panorama

    Presentation Overview: Show

    Break for lunch and poster viewing.

    Flash Journal Club
    Date: Sunday, July 23
    Time: 2:00 p.m. - 2:15 p.m.
    Room: Panorama
    • Frank Kramer, University Medical Center Goettingen, Germany

    Presentation Overview: Show

    A brief survey of recent, key papers in the Network Biology field.

    Multiple network-constrained regressions expand insights into influenza vaccination responses
    Date: Sunday, July 23
    Time: 2:15 p.m. - 2:30 p.m.
    Room: Panorama
    • Albert C. Shaw, Yale School of Medicine, United States
    • Steven H. Kleinstein, Yale School of Medicine, United States
    • Sui Tsang, Yale School of Medicine, United States
    • Barbara Siconolfi, Yale School of Medicine, United States
    • Samit R. Joshi, Yale School of Medicine, United States
    • Heidi Zapata, Yale School of Medicine, United States
    • Jean Wilson, Yale School of Medicine, United States
    • Stefan Avey, Yale School of Medicine, United States
    • Subhasis Mohanty, Yale School of Medicine, United States

    Presentation Overview: Show

    Systems immunology leverages recent technological advancements that enable broad profiling of the immune system to better understand the response to infection and vaccination, as well as the dysregulation that occurs in disease. An increasingly common approach to gain insights from these large-scale profiling experiments involves the application of statistical learning methods to predict disease states or the immune response to perturbations. However, the goal of many systems studies is not to maximize accuracy, but rather to gain biological insights. The predictors identified using current approaches can be uninterpretable or present only one of many equally predictive models, leading to a narrow understanding of the underlying biology.

    Here we show that incorporating prior biological knowledge within a logistic modeling framework by using network-level constraints on transcriptional profiling data significantly improves interpretability. Moreover, incorporating different types of biological knowledge produces models that highlight distinct aspects of the underlying biology, while maintaining predictive accuracy. We propose a new framework, Logistic Multiple Network-constrained Regression (LogMiNeR), and apply it to understand the mechanisms underlying differential responses to influenza vaccination. While standard logistic regression approaches were predictive, they were minimally interpretable. Incorporating prior knowledge using LogMiNeR led to models that were equally predictive yet highly interpretable. In this context, B cell-specific genes and mTOR signaling were associated with an effective vaccination response in young adults. Overall, our results demonstrate a new paradigm for analyzing high-dimensional immune profiling data in which multiple networks encoding prior knowledge are incorporated to improve model interpretability.

    A New Method to Study the Change of miRNA-mRNA Interactions Due to Environmental Exposures
    Date: Sunday, July 23
    Time: 2:30 p.m. - 2:45 p.m.
    Room: Panorama
    • Pei Wang, Icahn School of Medicine at Mount Sinai, United States
    • Susan L. Teitelbaum, Icahn School of Medicine at Mount Sinai, United States
    • Jia Chen, Icahn School of Medicine at Mount Sinai, United States
    • Nyan Win Khin, Icahn School of Medicine at Mount Sinai, United States
    • Maya Kappil, Icahn School of Medicine at Mount Sinai, United States
    • Kalpana Gopalakrishnan, Icahn School of Medicine at Mount Sinai, United States
    • Vasily N. Aushev, Icahn School of Medicine at Mount Sinai, United States
    • Francesca Petralia, Icahn School of Medicine at Mount Sinai, United States

    Presentation Overview: Show

    Motivation: Integrative approaches characterizing the interactions among different types of biological molecules have been demonstrated to be useful for revealing informative biological mechanisms. One such example is the interaction between microRNA (miRNA) and messenger RNA (mRNA), whose deregulation may be sensitive to environmental insult leading to altered phenotypes. The goal of this work is to develop an effective data integration method to characterize deregulation between miRNA and mRNA due to environmental toxicant exposures. We will use data from an animal experiment designed to investigate the effect of low-dose environmental chemical exposure on normal mammary gland development in rats to motivate and evaluate the proposed method.

    Results: We propose a new network approach - integrative Joint Random Forest (iJRF), which characterizes the regulatory system between miRNAs and mRNAs using a network model. iJRF is designed to work under the high-dimension low-sample-size regime, and can borrow information across different treatment conditions to achieve more accurate network inference. It also effectively takes into account prior information of miRNA-mRNA regulatory relationships from existing databases. When iJRF is applied to the data from the environmental chemical exposure study, we detected a few important miRNAs who regulated a large number of mRNA in the control group but not in the chemical exposure groups, suggesting disruptions of miRNA activities due to chemical exposure. Effects of chemical exposure on two miRNAs were further validated using breast cancer human cell lines.

    Genome-wide competing endogenous RNA networks highlight biomarkers in cancer
    Date: Sunday, July 23
    Time: 2:45 p.m. - 3:00 p.m.
    Room: Panorama
    • Christina Schultheiß, Saarland University, Germany
    • Sonja Kessler, Saarland University, Germany
    • Stephan Laggai, Saarland University, Germany
    • Azim Dehghani Amirabad, Saarland University, Germany
    • Marcel H. Schulz, Max Planck Institute for Informatics, Germany
    • Alexandra K. Kiemer, Saarland University, Germany
    • Dennis Kostka, University of Pittsburgh, United States
    • Markus List, Max Planck Institute for Informatics, Germany

    Presentation Overview: Show

    MicroRNAs (miRNAs) are small 19-22 nucleotide long molecules that facilitate the degradation of messenger RNA (mRNA) transcripts targeted via matching seed sequences. The competing endogenous RNA (ceRNA) hypothesis suggests that mRNAs that possess binding sites for the same miRNAs are in competition. This motivates the existence of so-called ​sponges​, i.e., genes that exert strong regulatory control via miRNA binding in a ceRNA interaction network. It is currently an unsolved problem how to estimate miRNA-mediated ceRNA interactions genome-wide. The most widely used approach considers miRNA and mRNA expression jointly measured for the same cell state. Several statistical methods have been proposed for determining ceRNA interaction strength using conditional mutual information or partial correlation, for instance. However, we identified three key limitations of existing approaches that prevent the construction of an accurate genome-wide ceRNA interaction network: (i) none of the existing methods considers the combinatorial effect of several miRNAs; (ii) due to the computational demand, the inference of a ceRNA interaction for all putative gene-miRNA-gene interactions in the human genome is prohibitive; (iii) an efficient strategy for determining the significance of inferred ceRNA interactions is missing, and thus important parameters of the system are neglected. To overcome these challenges, we developed a novel method called ​s​parse ​p​artial correlation ​on g​ene ​e​xpression (SPONGE) which is available as an R package. We reduce the computational complexity of constructing a genome-wide ceRNA interaction network in several steps. First, we consider only miRNA-gene interactions that are either predicted or experimentally validated. Second, we retain only miRNA-gene interactions that have a negative coefficient in a regularized regression model. Third, instead of each gene-miRNA-gene triplet, we compute a single sensitivity correlation (correlation - partial correlation) for each gene-gene pair given all shared miRNAs that pass the above filter as putative regulators. Finally, we derived the first mathematical formulation to simulate the null distribution of the process for different parameters of the system: number of miRNAs, correlation between genes and sample number. Our formulation enables the computation of empirical p-values for the statistic in a very efficient manner, an order of magnitude faster than previous methods. Analyses revealed that previous studies have underestimated the effect of these parameters in their network inference. In a first in-depth study we used SPONGE on data from The Cancer Genome Atlas to build a genome-wide model of ceRNA regulation in liver cancer. Network centrality measures revealed known and novel sponges, many of which are survival markers in liver cancer. We conducted additional experiments to validate selected ceRNA interactions in liver cancer. Our results highlight the relevance of ceRNA network inference for clinical research. In summary, we present the first method to solve the computationally demanding task of reporting significant ceRNA interactions efficiently on a genome-wide scale. Beyond ceRNAs, this method is well suited to infer other types of regulatory interactions such as transcription factor regulation.

    Using network analysis to identify a new key set of Parkinson’s Disease associated gene
    Date: Sunday, July 23
    Time: 3:00 p.m. - 3:15 p.m.
    Room: Panorama
    • Colin Mclean, University of Edinburgh, United Kingdom
    • Oksana Sorokina, University of Edinburgh, United Kingdom
    • Katharina F Heil, University of Edinburgh / KTH Stockholm, United Kingdom
    • J Douglas Armstrong, University of Edinburgh, United Kingdom

    Presentation Overview: Show

    Parkinson’s Disease (PD), the second most common neurodegenerative disease in the western world shows a high degree of genetic complexity with many implicated factors but few clear causal links. Ongoing research has identified a growing number of dysfunctions linked to the disease. Many biochemical pathways, such as the mitochondrial respiratory chain, the ubiquitination system and vesicle cycling are often affected and have been studied in detail. To fully understand PD, find the causative mechanisms and identify potential drug-targets it is crucial to analyse how these pathways interlink. Omics data availability is increasing on a fast pace, but has not yet been systematically analysed in the context of PD. After curating available disease-gene association data based on mutations and public annotations a comprehensive set of 420 PD associated genes was obtained. The synapse is one of the key affected areas in PD. A recent review of published synaptic proteomic studies (in preparation), including presynapse and postsynapse specific data identified around 6,500 proteins. Analysis of protein coverage in a specific number of analysed studies leads to the consensus synaptic proteome, including around 2,500 proteins (with minimum coverage of three studies). Curated, direct, human protein-protein-interaction data (including data from BIOGRID, dip and IntAct) was used to generate a protein-protein interaction network (PPIN) model of the 2,500 synaptic proteins. In this study we looked in detail at the intersection of the two datasets, introduced above. In particular we looked at the distribution of PD genes in subregions of the synaptic PPINs. The networks were clustered using different algorithms including: 1) modularity based approaches, such as the agglomerative fast-greedy and Random Walk, 2) divisive spectral modularity as well as 3) hierarchical agglomerative lourvain algorithm and 4) random walk based infomap. The analysis of different network parameters such as modularity, clustering coefficient, node degree and centrality as well as community robustness amongst others, gives different insights into the data structure and disease implication. PD associated genes were located in the networks and hypergeometric testing identified significantly enriched clusters. Functional Gene Ontology analysis determined “clathrin mediated endocytosis” as well as “synaptic vesicle cycling” as key affected functions of all genes in PD-associated clusters, irrelevant of the supplied algorithm. Figure 1 visualises the analytical approach and shows a network of the consensus PPIN highlighting PD-enrichment and associated functions. Based on this approach a new list of PD-associated genes was identified. This set of 56 genes gives new insights into underlying triggers and overall functionalities, affected by and shaping the PD genotype and inducing its phenotype. Further analysis is required to pin down potential biomarkers or drug-targets, which can speed up further efforts.

    Reconstruction and signal propagation analysis of the syk signaling network in breast cancer cells
    Date: Sunday, July 23
    Time: 3:15 p.m. - 3:30 p.m.
    Room: Panorama
    • Peter Coopman, INSERM, France
    • Ovidiu Radulescu, University of Montpellier, France
    • Romain Larive, Université de Montpellier, France
    • Aurélien Naldi, Université de Montpellier, France

    Presentation Overview: Show

    The ability to build in-depth cell signaling networks from vast experimental data is a major objective of computational biology. The Syk protein tyrosine kinase, a well- characterized key player in immune cell signaling, was surprisingly first shown by our group to exhibit an onco-suppressive function in mammary epithelial cells and corroborated by many other studies, but the molecular mechanisms of this function remain largely unsolved. Based on existing proteomic data, we report here the generation of an interaction-based network of signaling pathways controlled by Syk in breast cancer cells. Pathway enrichment of the Syk targets previously identified by quantitative phospho-proteomics indicated that Syk is engaged in cell adhesion, motility, growth and death. Using the components and interactions of these pathways, we bootstrapped the reconstruction of a comprehensive network covering Syk signaling in breast cancer cells. To generate in silico hypotheses on Syk signaling propagation, we developed a method allowing to rank paths between Syk and its targets. We first annotated the network according to experimental datasets. We then combined shortest path computation with random walk processes to estimate the importance of individual interactions and selected biologically relevant pathways in the network. Molecular and cell biology experiments allowed to distinguish candidate mechanisms that underlie the impact of Syk on the regulation of cortactin and ezrin, both involved in actin-mediated cell adhesion and motility. The Syk network was further completed with the results of our biological validation experiments. The resulting Syk signaling sub-networks can be explored via an online visualization platform.

    Network based approach for analysis of cell heterogeneity and immune polarization in tumor microenvironment from single-cell data
    Date: Sunday, July 23
    Time: 3:30 p.m. - 3:35 p.m.
    Room: Panorama
    • Inna Kuperstein, Institut Curie, France
    • Emmanuel Barillot, Institut Curie, France
    • Urszula Czerwińska, Institut Curie, France
    • Maria Kondratova, Institut Curie, France
    • Andrei Zinovyev, Institut Curie, France

    Presentation Overview: Show

    Tumor microenvironment (TME) plays important and, sometimes, opposite roles in tumor evolution. We developed and applied new network-based approach to analyse both non-immune (cancer associated fibroblasts (CAF)) and immune (macrophages (Mph) and natural killers (NK)) components of this multicellular system. Our approach is based on reconstructing the cell-type specific networks of molecular interactions involved in tumor-TME interaction according to the methodology proposed in (Kuperstein et al. 2015; Kondratova et al. 2017). We created cell-specific CAFs network map composed of 681 objects and 585 reactions, the first integrated network representing together pathways involved in fibroblast activation in cancer. The CAFs map has modular structure covering all main functions of CAFs in tumor among which the interactions of CAFs with extracellular matrix; signaling coordinating involvment of CAFs in tumor growth; interactions of fibroblasts with immune system. In addition, there are functional modules responsible for fibroblast activation (pro-tumor activity) and inhibition (anti-tumor activity) along with the metabolic functional module that includes the main pathways involved in metabolic reprogramming of CAFs and ROS production. In addition, we created meta-map of innate immune response in cancer composed of 1476 objects and 1085 reactions. We constructed signalling maps of macrophages, dendritic cells, myeloid-derived suppressor cells, natural killers, neutrophils and mast cells. These cell-specific maps integrated together and updated by interactions and crosstalks between them and the map of tumor cell, gave rise to a seamless comprehensive meta-map of innate immune response in cancer, demonstrating signalling responsible for anti- and pro-tumour activities of innate immunity system as a whole. It is a ‘geographical-like’ hierarchically organized meta-map with functional ‘zones’, namely, signalling mechanisms contributing to anti-tumor or pro-tumor immune phenotypes. Finally, we applied these network maps for identification of molecular mechanisms regulating cell reprogramming in several innate immune cell types. We applied unsupervised statistical methods for decomposition of single cell RNASeq data for fibroblasts, natural killers and macrophages from melanoma (Tirosh et al. 2016). Independent component analysis-based methodology (Biton et al., 2014) decomposed single cell transcriptiome data into components and highlighted patterns associated with different cell populations in TME. This unsupervised analysis highlighted expression patterns associated with different cell sub-types. Analysis and interpretation of the expression patterns in the context of innate immunity network map revealed characteristic functional properties visualizing signalling associated with anti- and pro-tumor activity in each cell sub-type across fibroblasts, natural killers and macrophages populations in melanoma.

    A PRObabilistic Pathway Score (PROPS) for Classification with Applications to Inflammatory Bowel Disease
    Date: Sunday, July 23
    Time: 3:35 p.m. - 3:40 p.m.
    Room: Panorama
    • Lovisa Afzelius, Pfizer, Inc, United States
    • Joshua R. Korzenik, Brigham & Women's Hospital, Harvard Medical School, United States
    • Scott B. Snapper, Boston Children's Hospital, Harvard Medical School, United States
    • Mateusz Maciejewski, Pfizer, Inc, United States
    • Russ B. Altman, Stanford University, United States
    • William Gordon, Pfizer Inc, United States
    • Christoph Brockel, Pfizer, Inc, United States
    • Lichy Han, Stanford University, United States

    Presentation Overview: Show

    Gene-based supervised machine learning classification models have become a widely-used tool in biomedicine to help diagnose diseases, determine prognosis, and predict drug response. However, many of these classifiers are sensitive to noise and measurement error, which often results in low reproducibility in external validation sets. For complex diseases that are heterogeneous and polygenic, such as inflammatory bowel disease (IBD), these classifiers are further limited by their inability to capture varying combinations of genes that result in the same clinical phenotype across different patients and datasets. Pathway-based classification can overcome these challenges by aggregating genes into robust, pathway-based features which represent underlying biological mechanisms. In doing so, pathway-based classification uses biological regularization to combat overfitting and to capture biological mechanisms inherent to the phenotype of interest. In this work, we present a novel pathway-based approach, PRObabilistic Pathway Score (PROPS), which aggregates genes to calculate individualized pathway-based scores for classification. Unlike previous individualized pathway-based classification methods, which are based on gene sets, we use probabilistic graphical models to incorporate gene interactions and model each pathway as a directed network. We apply our method to differentiating the two main types of IBD, ulcerative colitis (UC) and Crohn’s disease (CD), which are two highly similar complex diseases. Using five IBD datasets, we compared our method against eight other feature engineering methods: four gene-based methods, including a previously published five gene IBD classifier, and four alternative pathway-based methods. We trained on the largest dataset and independently evaluated using the four remaining datasets. We constructed 100 random forest models for each feature set and evaluated each method using the median area under the receiver-operating characteristic curve (AUC). We demonstrate superior classification performance (Figure 1) while providing biological insight into the top pathways separating CD form UC.

    Comprehensive analysis of high-throughput screens with HiTSeekR – From RNAi, CRISPR/Cas9, miRNA and small compound screens to targeted signalling pathways
    Date: Sunday, July 23
    Time: 3:40 p.m. - 3:45 p.m.
    Room: Panorama
    • Jan Mollenhauer, NanoCAN - University of Southern Denmark, Denmark
    • Marc Rehmsmeier, Humboldt Universität zu Berlin, Germany
    • Steffen Schmidt, Roche Innovation Center Copenhagen, Denmark
    • Jan Baumbach, University of Southern Denmark, Denmark
    • Qihua Tan, Odense University Hospital, Denmark
    • Helle Christiansen, Roche Innovation Center Copenhagen A/S, Denmark
    • Markus List, Max Planck Institute for Informatics, Germany

    Presentation Overview: Show

    High-throughput screening (HTS) is an indispensable tool for drug (target) discovery that currently lacks user-friendly software tools for the robust identification of putative hits from HTS experiments and for the interpretation of these findings in the context of systems biology. Limited resources only allow for a few putative hits (active samples) to be considered for follow-up experiments. In contrast, systems biomedicine analysis is suited to identify targeted pathways more efficiently based on the entire data set. We developed HiTSeekR as a one-stop solution for chemical compound screens, siRNA knock-down and CRISPR/Cas9 knock-out screens, as well as microRNA inhibitor and -mimics screens. For each screen type, HiTSeekR enables the user to extract a list of (putative) target genes that can be subjected to gene set and network enrichment analysis. The latter is particularly suited for drug target discovery as it allows for extracting novel and disease related functional modules from biological interaction networks. We will present three use cases that demonstrate how HiTSeekR may be used to fully exploit HTS screening data in quite heterogeneous contexts to generate novel hypotheses for follow-up experiments: (1) a genome-wide RNAi screen to uncover modulators of TNF-alpha, (2) a combined siRNA and miRNA mimics screen on vorinostat resistance and (3) a small compound screen on KRAS synthetic lethality. HiTSeekR is the first approach to close the gap between raw data processing, network enrichment and wet lab target generation for various HTS screen types.

    NetRep: a scalable permutation approach for assessing replication and preservation of network modules in large datasets
    Date: Sunday, July 23
    Time: 3:45 p.m. - 3:50 p.m.
    Room: Panorama
    • Gad Abraham, The University of Melbourne, Australia
    • Kathryn Holt, The University of Melbourne, Australia
    • Liam Fearnley, The University of Melbourne, Australia
    • Stephen Watts, The University of Melbourne, Australia
    • Scott Ritchie, The University of Melbourne, Australia
    • Michael Inouye, The University of Melbourne, Australia

    Presentation Overview: Show

    Network inference techniques are widely used to identify and characterize complex relationships between genomic, transcriptomic, metabolomics, and proteomic data measured by high-throughput platforms. Network modules—topologically distinct groups of edges and nodes within these networks—that are preserved across datasets can reveal common features of cell-types, tissues, and organisms. Many statistics for assessing the preservation of network module topology have been developed, however, testing their significance requires heuristics [1]. Consequently, these statistics cannot be adjusted for multiple testing, which is important as the number of modules and datasets undergoing module preservation increases with large multi-omic datasets becoming increasing common and openly available [2]. We introduced NetRep, a fast, scalable, and statistically rigorous method for assessing module preservation through permutation testing without assuming data are normally distributed. NetRep produces unbiased p values and can distinguish between true and false positives during multiple hypothesis testing. NetRep is published as an R package on CRAN (https://cran.r- project.org/package=NetRep). In a recent study, we demonstrated current statistics for assessing module preservation are systematically biased and produce skewed P–values [3]. We used NetRep to quantify preservation of gene coexpression modules across murine brain, liver, adipose, and muscle tissues. Complex patterns of multi-tissue preservation were revealed, including a liver-derived housekeeping module that displayed adipose- and muscle- specific association with body weight. Finally, we demonstrate the broad applicability of NetRep by quantifying preservation of bacterial networks in gut microbiota between men and women.

    An R package for diffusion algorithms in biological networks
    Date: Sunday, July 23
    Time: 3:50 p.m. - 3:55 p.m.
    Room: Panorama
    • Alexandre Perera, Universitat Politècnica de Catalunya, Spain
    • Alfonso Buil, Mental Health Center Sct. Hans, Denmark
    • Wesley Thompson, Mental Health Center Sct. Hans, Denmark
    • Sergio Picart-Armada, Universitat Politècnica de Catalunya, Spain

    Presentation Overview: Show

    The guilt-by-association principle states that interacting molecular entities are prone to share biological properties and functions. This motivates the usage of the network structure of annotated knowledge as contextual data: labels can be propagated and quantitative measures can be smoothed over interacting molecular entities. The success of diffusion in biological networks reaches a wide range of applications, for instance drug discovery, gene-disease association, protein function prediction and genome-wide association studies hit prioritisation. Despite the success of such techniques, each individual application uses a particular configuration for the diffusion process, so benchmarking between approaches at the methodological level becomes difficult. To that end, we have integrated both widely used and less common diffusion kernels and scores in an R package that eases the comparison among them. Our package has been applied to a yeast interactome with protein labels. The implemented diffusion scores have been compared in terms of predictive power using ad-hoc built-in functions, showing that recently proposed statistically normalised scores outperform their classical counterparts in this scenario.

    Coffee Break
    Date: Sunday, July 23
    Time: 4:00 p.m. - 4:30 p.m.
    Room: Panorama

      Presentation Overview: Show

      Break for coffee and poster viewing.

      The STRING app: bringing quality-controlled protein-protein and protein-chemical networks into Cytoscape
      Date: Sunday, July 23
      Time: 4:30 p.m. - 4:35 p.m.
      Room: Panorama
      • Jan Gorodkin, University of Copenhagen, Denmark
      • John H. Morris, University of California, San Francisco, United States
      • Nadezhda T. Doncheva, University of Copenhagen, Denmark
      • Lars Juhl Jensen, University of Copenhagen, Denmark

      Presentation Overview: Show

      Many high-throughput technologies such as mass spectrometry result in a long list of proteins associated with a certain condition or phenotype. The physical and functional associations among these proteins can provide important cellular context for understanding their role. Such association networks for 2000+ organisms can be obtained from the STRING and STITCH databases, which provide quality-controlled protein–protein and protein–chemical association, respectively. Standalone tools, such as the Cytoscape software framework, provide numerous features for integrating user data with molecular networks and for analyzing and visualizing these networks in various ways. However, making good use of these resources and combining them at the same time can be challenging for many users. Here, we present the STRING app, which allows users to directly import molecular networks from STRING and STITCH into Cytoscape for any list of proteins or compounds of interest. The resulting networks can, for example, be combined with the user’s own data to visualize regulation or analyzed for enriched functions, which are shown in a new data table within Cytoscape (see Fig. 1). The app also offers two further ways of querying for STRING interactions – for the top genes associated with a disease of interest according to the DISEASES database and for the top genes associated with a topic of interest as determined by text mining of all PubMed abstracts.

      HGPEC: a Cytoscape app for prediction of novel disease-gene and disease-disease associations and evidence collection based on a random walk on heterogeneous network
      Date: Sunday, July 23
      Time: 4:35 p.m. - 4:40 p.m.
      Room: Panorama
      • Duc-Hau Le, VINMEC Research Institute of Stem Cell and Gene Technology, Viet Nam

      Presentation Overview: Show

      Background: Finding gene-disease and disease-disease associations play important roles in the biomedical area and many prioritization methods have been proposed for this goal. Among them, approaches based on a heterogeneous network of genes and diseases are considered state-of-the-art ones, which achieve high prediction performance and can be used for diseases with/without known molecular basis. Results: Here, we developed a Cytoscape app, namely HGPEC, based on a random walk with restart algorithm on a heterogeneous network of genes and diseases. This app can prioritize candidate genes and diseases by employing a heterogeneous network consisting of a network of genes/proteins and a phenotypic disease similarity network. Based on the rankings, novel disease-gene and disease- disease associations can be identified. These associations can be supported with network- and rank- based visualization as well as evidences and annotations from biomedical data. A case study on prediction of novel breast cancer-associated genes and diseases shows the abilities of HGPEC. In addition, we showed prominence in the performance of HGPEC compared to other tools for prioritization of candidate disease genes. Conclusions: Taken together, our app is expected to effectively predict novel disease-gene and disease-disease associations and support network- and rank-based visualization as well as biomedical evidences for such the associations. Access the app at: https://sites.google.com/site/duchaule2011/bioinformatics-tools/hgpec

      Pathways on demand: automated reconstruction of human signaling networks
      Date: Sunday, July 23
      Time: 4:40 p.m. - 4:45 p.m.
      Room: Panorama
      • Anna Ritz, Reed College, United States
      • Christopher Poirel, RedOwl Analytics, United States
      • Allison Tegge, Virginia Tech, United States
      • Nicholas Sharp, Virginia Tech, United States
      • Kelsey Simmons, Virginia Tech, United States
      • Allison Powell, Virginia Tech, United States
      • Shiv Kale, Virginia Tech, United States
      • T. M. Murali, Virginia Tech, United States

      Presentation Overview: Show

      Signaling pathways are a cornerstone of systems biology. Several databases store high-quality represen- tations of these pathways. Despite painstaking and manual curation, these databases remain incomplete. To complement manual curation, we sought to develop a computational approach to automatically reconstruct signaling pathways. We conceptualized the problem as follows: given as input only the receptors and the transcription factors/regulators (TRs) in a specific signaling pathway, can we analyze a background network of molecular interactions (the interactome) to recover the pathway with high accuracy? Our algorithm, named PATHLINKER, efficiently computes multiple short paths within the interactome that connect the receptors to the TRs in a pathway [5]. Specifically, PATHLINKER computes the k highest scoring receptor-to-TR paths, where k is a user-specified parameter, using a novel integration of Yen’s algorithm [8] with the A* heuristic. Even with k = 20, 000, PATHLINKER takes only 30 seconds to run on average per pathway in an interactome with 12,046 proteins and 152,094 interactions. Thus, this technique is capable of handling the complexity of human interaction networks. We have developed a Cytoscape app that has been downloaded over 2,100 times in 16 months [2]. We used PATHLINKER to accurately reconstruct a comprehensive set of signaling pathways from the NetPath databases [5]. We compared PATHLINKER with six other state-of-the-art network-based algorithms, including shortest path (SHORTESTPATHS, BOWTIEBUILDER [6]), random walk with restarts (RWR [3]), network flow (RESPONSENET [7]), Steiner forest (PCSF [1]), ANAT [9], and a greedy seed-based method (IPA [4]). PATHLINKER achieved higher precision and recall than these algorithms (Figure 1(left)). Only PATHLINKER, RWR, and IPA could achieve recall values larger than 0.1. PATHLINKER was the only method that could control the size of the reconstruction while ensuring that receptors were connected to transcription factors in the output network (Figure 1(right)). In more recent work, we have extended these results to a much larger set of NetPath and to KEGG pathways. PATHLINKER’s reconstruction of the Wnt pathway identified CFTR, an ABC class chloride ion channel transporter, as a novel intermediary that facilitates the signaling of Ryk to Dab2, which are known com- ponents of this pathway. We experimentally validated this prediction in HEK293 cells, showing that the Ryk–CFTR–Dab2 path is a novel amplifier of β-catenin signaling specifically in response to Wnt 1/2/3/3a. PATHLINKER’s success in reconstructing NetPath and KEGG pathways point to its applicability for complementing manual curation. PATHLINKER is a promising approach for prioritizing proteins and inter- actions for experimental study, as illustrated by its discovery of a novel pathway in Wnt/β-catenin signaling.

      Predicting Novel Metabolic Pathways through Subgraph Mining
      Date: Sunday, July 23
      Time: 4:45 p.m. - 5:00 p.m.
      Room: Panorama
      • Aravind Sankar, Indian Institute of Technology Madras, India
      • Sayan Ranu, Indian Institute of Technology Delhi, India
      • Karthik Raman, Indian Institute of Technology Madras, India

      Presentation Overview: Show

      The ability to predict pathways for biosynthesis of metabolites is very important in metabolic engineering. It is possible to mine the repertoire of biochemical transformations from reaction databases, and apply the knowledge to predict reactions to synthesise new molecules. However, this usually involves a careful understanding of the mechanism and the knowledge of the exact bonds being created and broken. There is clearly a need for a method to rapidly predict reactions for synthesising new molecules, which relies only on the structures of the molecules, without demanding additional information such as thermodynamics or hand-curated information such as atom-atom mapping, which are often hard to obtain accurately. We here describe a robust method based on subgraph mining, to predict a series of biochemical transformations, which can convert between two (even previously unseen) molecules. We first describe a reliable method based on subgraph edit distance to map reactants and products, using only their chemical structures. Having mapped reactants and products, we identify the reaction centre and its neighbourhood, the reaction signature, and store this in a reaction rule network. This novel representation enables us to rapidly predict pathways, even between previously unseen molecules. We also propose a heuristic that predominantly recovers natural biosynthetic pathways from amongst hundreds of possible alternatives, through a directed search of the reaction rule network, enabling us to provide a reliable ranking of the different pathways. Our approach scales well, even to databases with >100,000 reactions.

      NetBio Keynote: Pathway analysis of genomics data. From correlation to causation to drug discovery.
      Date: Sunday, July 23
      Time: 5:00 p.m. - 5:45 p.m.
      Room: Panorama
      • Gary Bader, University of Toronto, Canada

      Presentation Overview: Show

      Invited keynote presentation.

      NetBio Closing
      Date: Sunday, July 23
      Time: 5:45 p.m. - 6:00 p.m.
      Room: Panorama

        Presentation Overview: Show

        Wrap-up and community topics.