Search

This report documents the program and the outcomes of Dagstuhl Seminar 19061 "Visual Analytics of Multilayer Networks Across Disciplines". Networks, used to understand systems, often contain multiple types of nodes and/or edges. They are often flattened to a single network, even though real-world systems are more accurately modelled as a set of interacting networks, or layers, with different node and edge types. These are so-called multilayer networks. These networks are studied by researchers both in network visualization and in complex systems -- the domain from which the concept of multilayer networks has recently emerged. Moreover, researchers in various application domains study these systems, e.g. biology, digital humanities, sociology and journalism. These research areas have shown parallel individual developments. Therefore, one of the aims of the seminar was to bring together an interdisciplinary community of researchers and practitioners of different disciplines. This interdisciplinary community discussed existing solutions, open challenges and future research directions for visual analytics of multilayer networks across disciplines. The seminar was attended by researchers from information visualization, visual analytics, complex systems and application domains. The application domains covered digital humanities, social sciences, biological sciences, and in public health research (25% of attendees were from these fields). The seminar not only provided multiple application domains for the visualization experts, but also also provided the domains experts with different groups of visualization experts in breakouts sessions, to expose them to multiple approaches to solving their problems. Building on this close working relationship between the visualization and domain experts, working groups were defined to determine which are the important challenges for multilayer network visualization. A number of sub-topics were identified that require further research: A

Visualization of biological mechanisms by means of pathway graphs is necessary to better understand the often complex underlying system. Manual layout of such pathways or maps of knowledge is a difficult and time consuming process. Node duplication is a technique that makes layouts with improved readability possible by reducing edge crossings and shortening edge lengths in drawn diagrams. In this article we propose an approach using Machine Learning (ML) to facilitate parts of this task by training a Support Vector Machine (SVM) with actions taken during manual biocuration. Our training input is a series of incremental snapshots of a diagram describing mechanisms of a disease, progressively curated by a human expert employing node duplication in the process. As a test of the trained SVM models, they are applied to a single large instance and 25 medium-sized instances of hand-curated biological pathways. Finally, in a user validation study, we compare the model predictions to the outcome of a node duplication questionnaire answered by users of biological pathways with varying experience. We successfully predicted nodes for duplication and emulated human choices, demonstrating that our approach can effectively learn human-like node duplication preferences to support curation of pathway diagrams in various contexts.

This report documents the program and the outcomes of Dagstuhl Seminar 19061 "Visual Analytics of Multilayer Networks Across Disciplines". Networks, used to understand systems, often contain multiple types of nodes and/or edges. They are often flattened to a single network, even though real-world systems are more accurately modelled as a set of interacting networks, or layers, with different node and edge types. These are so-called multilayer networks. These networks are studied by researchers both in network visualization and in complex systems -- the domain from which the concept of multilayer networks has recently emerged. Moreover, researchers in various application domains study these systems, e.g. biology, digital humanities, sociology and journalism. These research areas have shown parallel individual developments. Therefore, one of the aims of the seminar was to bring together an interdisciplinary community of researchers and practitioners of different disciplines. This interdisciplinary community discussed existing solutions, open challenges and future research directions for visual analytics of multilayer networks across disciplines. The seminar was attended by researchers from information visualization, visual analytics, complex systems and application domains. The application domains covered digital humanities, social sciences, biological sciences, and in public health research (25% of attendees were from these fields). The seminar not only provided multiple application domains for the visualization experts, but also also provided the domains experts with different groups of visualization experts in breakouts sessions, to expose them to multiple approaches to solving their problems. Building on this close working relationship between the visualization and domain experts, working groups were defined to determine which are the important challenges for multilayer network visualization. A number of sub-topics were identified that require further research: A

BACKGROUND: Recent advances in high-throughput sequencing allow for much deeper exploitation of natural and engineered microbial communities, and to unravel so-called "microbial dark matter" (microbes that until now have evaded cultivation). Metagenomic analyses result in a large number of genomic fragments (contigs) that need to be grouped (binned) in order to reconstruct draft microbial genomes. While several contig binning algorithms have been developed in the past 2 years, they often lack consensus. Furthermore, these software tools typically lack a provision for the visualization of data and bin characteristics. RESULTS: We present ICoVeR, the Interactive Contig-bin Verification and Refinement tool, which allows the visualization of genome bins. More specifically, ICoVeR allows curation of bin assignments based on multiple binning algorithms. Its visualization window is composed of two connected and interactive main views, including a parallel coordinates view and a dimensionality reduction plot. To demonstrate ICoVeR's utility, we used it to refine disparate genome bins automatically generated using MetaBAT, CONCOCT and MyCC for an anaerobic digestion metagenomic (AD microbiome) dataset. Out of 31 refined genome bins, 23 were characterized with higher completeness and lower contamination in comparison to their respective, automatically generated, genome bins. Additionally, to benchmark ICoVeR against a previously validated dataset, we used Sharon's dataset representing an infant gut metagenome. CONCLUSIONS: ICoVeR is an open source software package that allows curation of disparate genome bins generated with automatic binning algorithms. It is freely available under the GPLv3 license at https://git.list.lu/eScience/ICoVeR . The data management and analytical functions of ICoVeR are implemented in R, therefore the software can be easily installed on any system for which R is available. Installation and usage guide together with the example files ready to be vis

Our growing knowledge about various molecular mechanisms is becoming increasingly more structured and accessible. Different repositories of molecular interactions and available literature enable construction of focused and high-quality molecular interaction networks. Novel tools for curation and exploration of such networks are needed, in order to foster the development of a systems biology environment. In particular, solutions for visualization, annotation and data cross-linking will facilitate usage of network-encoded knowledge in biomedical research. To this end we developed the MINERVA (Molecular Interaction NEtwoRks VisuAlization) platform, a standalone webservice supporting curation, annotation and visualization of molecular interaction networks in Systems Biology Graphical Notation (SBGN)-compliant format. MINERVA provides automated content annotation and verification for improved quality control. The end users can explore and interact with hosted networks, and provide direct feedback to content curators. MINERVA enables mapping drug targets or overlaying experimental data on the visualized networks. Extensive export functions enable downloading areas of the visualized networks as SBGN-compliant models for efficient reuse of hosted networks. The software is available under Affero GPL 3.0 as a Virtual Machine snapshot, Debian package and Docker instance at http://r3lab.uni.lu/web/minerva-website/. We believe that MINERVA is an important contribution to systems biology community, as its architecture enables set-up of locally or globally accessible SBGN-oriented repositories of molecular interaction networks. Its functionalities allow overlay of multiple information layers, facilitating exploration of content and interpretation of data. Moreover, annotation and verification workflows of MINERVA improve the efficiency of curation of networks, allowing life-science researchers to better engage in development and use of biomedical knowledge repositories.