COMBINE 2018 - Abstracts

Here, we present a new method for the visualization of time-course metabolomics data within the context of a metabolic network map. We demonstrate the utility of this method by examining previously published data for two cellular systems—the human platelet and erythrocyte under cold storage for use in transfusion medicine. Our approach allowed for new insights into the state of the platelet metabolome during storage, most notably that nicotinamide secretion mirrors that of hypoxanthine and might, therefore, reflect similar pathway usage. These results are compiled into descriptive animation videos that demonstrate the dynamic metabolite concentrations across the entire metabolic network, as well as specific focus on individual pathways.

This method can be applied to any system with data and a metabolic map to help visualize and understand physiology at the network level. More broadly, we hope that our approach will provide the blueprints for new visualizations of other longitudinal -omics data types.

Sub-SBML provides functions to create subsystems for SBML models in order to use various utility functions to edit the model (such as renaming different lists of model components, creating multiple components at once, integrations for different simulator options, simulating variable species amount in a model etc). It is often desired in modeling of compounded biomolecular systems to combine multiple smaller models for different modules. The primary features of Sub-SBML allow for modeling of such systems by creating subsystems for different models and then combining these subsystems as desired, while also modeling the interactions among the models. A list of subsystems may be combined by specifying a list of species which would be shared in the final combined model. The toolbox also provides functions to model the cases when it is desired that all species that have the same name be a single combined species in the final model.

Sub-SBML can also be used to model compartments in SBML for modeling the transport of species between different models. A subsystem object in Sub-SBML consists of an SBML model that can have one compartment to hold all the species of that model. Multiple subsystems may be placed inside a “system”. A system object in Sub-SBML acts as a container for subsystems. Subsystems may be defined to be internal to a system, external to a system, or as membranes to a system. Hence, an SBML model of a system would consist of two compartments – internal and external. The membrane subsystem will also have these two compartments and reactions modeling the transport of species across it. A system model can be generated that would give a combined model of all subsystems that are internal, external and set as membrane to this system, hence modeling the transport across this system container. Similarly, multiple system objects can be modeled and interactions across multiple systems may be modeled as well.

We consider an example of IPTG transport across synthetic vesicles when α-hemolysin is expressed inside the vesicle, which creates pores in the membrane to allow transport of IPTG molecules from the environment to inside the vesicle. We demonstrate the use of Sub-SBML to model this system in SBML. We also consider other synthetic biological circuits to demonstrate input-output species interactions among subsystems and to obtain combined SBML models.

We have developed ngBioCAD, a computer-aided design (CAD) tool for synthetic biology built on the SBOL2 and SBOL Visual standards. ngBioCAD is entirely Web-based, and uses an intuitive “drag and drop” approach to design common to CAD tools outside of synthetic biology. Features supported by ngBioCAD include hierarchical top-down design, specification of interactions, and sequence editing. The authors hope that ngBioCAD will both promote the adoption of the SBOL2 standard and serve as a useful tool for the design stage of the synthetic biology lifecycle.

SBMLme is an extension of current model encoding standards that enables SBML representations of ME-models that were reconstructed using COBRAme [2]. A prototype of the extension has been created in Java together with a standalone, bi-directional converter, between this extension and COBRAme's model storage format. The converter showed that SBMLme could fully and correctly encode a COBRAme model.

The use of SBMLme enables ME-models to be shared more efficiently, and a wider variety of tools to access ME-models. This should promote the propagation of ME-models. SBMLme may be used as a proof-of-concept for an official SBML package for ME-models. The standardization process requires ongoing discussion, the identification of a consensus SBML draft standard, and the implementation of the existing validation rules.

SBMLme is freely available at https://github.com/draeger-lab/SBMLme under the terms of the MIT license.

References:

[1] Ines Thiele, Ronan M. T. Fleming, Richard Que, Aarash Bordbar, Dinh Diep, and Bernhard O. Palsson. Multiscale Modeling of Metabolism and Macromolecular Synthesis in E. coli and Its Application to the Evolution of Codon Usage. PLOS ONE, 7(9):1-18, 09 2012.

[2] Colton J Lloyd, Ali Ebrahim, Laurence Yang, Zachary Andrew King, Edward Catoiu, Edward J O'Brien, Joanne K Liu, and Bernhard O Palsson. COBRAme: A Computational Framework for Models of Metabolism and Gene Expression. bioRxiv, 2017.

We aimed to get a deeper understanding of the complex relationships between pathogen, host, and microbiome. To predict pathogen expansion, gut colonization and infection outcome we employed a powerful measure of systems biology, i.e., the development of a computational model.For implementation and challenge of the model, oral mouse infection experiments with the enteropathogen Yersinia enterocolitica (Ye) were used. Our model calculates the bacterial population dynamics during gastrointestinal infection and accounts for specific pathogen characteristics, the host immune capacity and colonization resistance mediated by the endogenous microbiome. First, we performed model parameter optimization based on the experimental data we obtained by the infection of a healthy host. Afterward, we challenged our model by adopting scenarios where either a microbiome was lacking (mimicking antibiotic treatment of patients), or where the immune response was partially impaired. The predicted Ye population dynamics based on these scenarios could be approved in experimental mouse infections.

Our model provides new hypotheses about the roles of host- and pathogen-derived factors within this complex interplay and might be useful for future development of personalized infection prevention and treatment strategies.

By using the FAIRDOMHub, researchers can achieve more effective exchange with geographically distributed collaborators during projects, ensure results are sustained and preserved and generate reproducible publications that adhere to the FAIR guiding principles of data stewardship.

Functionality to upload and download COMBINE archives to and from SynBioHub has been implemented. SBOL files in the archive are inserted into a general-purpose SBOL Collection created for the archive. Other files are uploaded to the filestore, and linked to the SBOL Collection using the new SBOL Attachment TopLevel object. This functionality was validated using the iBioSim GDA tool and libSBOLj's SynBioHubFrontend.

This presentation will focus on SBOLExplorer, a system that is used to provide intuitive search within the SynBioHub genetic part repository. SynBioHub integrates genetic construct data from various sources and transforms and stores this data in the Synthetic Biology Open Language (SBOL), a standardized data model. By tackling the intricate data mining and data infrastructure problems associated with large scale unstructured and noisy data, the search, transformation, and storage of SynBioHub data can be enhanced. In particular, this presentation will focus on improving the usability of SynBioHub's search capabilities. By clustering SynBioHub's genetic parts into many derived collections, duplicate parts can be merged and diverse results can be shown. From there, a graph analysis algorithm can be designed to rank collections of parts by popularity and usefulness. Finally, data infrastructure challenges relating to indexing, storing, and serving need to be solved. The end goal is to integrate these findings into SynBioHub's data representation, search functionality, and data infrastructure.

To this end many grassroots standards for data, models and their metadata have been defined by the scientific communities and are driven by standardization initiatives such as the Computational Modeling in Biology Network (COMBINE). To fill gaps in domains that currently have a lack of standardization, as for example the field of multicellular and multiscale modelling, we define together with our partners new modelling standards such as MultiCellML as a new exchange format for multicellular models.

Moreover, for integration of data and models standards have to be harmonized to be interoperable and allow interfacing between the datasets. We drive and lead the definition of novel standards of the International Organization for Standardization (ISO) in the technical ISO committee for biotechnology standards (ISO/TC 276) in order to define a framework and guideline for community standards and their application. With our activities we aim at enhancing the interoperability of community standards for life science data and models and therefore facilitating complex and multiscale data integration and model building with heterogenous data gathered across the domains. This is accompanied by activities that build a bridge between stakeholders and by our activities to develop the means and channels for transferring information about standards between them, such as the NormSys registry for modelling standards (http://normsys.h-its.org).

Public model repositories, such as the Physiome Model Repository (PMR)1 and BioModels Database2, are vital resources for accessing computational models of biological processes. There is broad recognition that these resources provide value for scientists aiming to build from others’ works. Both repositories are manually curated: Curators identify a publication with a model, and then work to develop and establish reproducibility of the code. These methods do not scale well with the pace of publication. Without these third-party curators, what percentage of models described in the literature are reproducible? More fundamentally, we investigate what percentage of publications about models include some access to, or information about the model code.

Methods: In order to elucidate the status of model availability and reproducibility in literature, we conducted a scoping review to characterize computational physiology models in literature. We looked at whether or not (a) the model code is available, (b) the modeling language used is stated, and (c) the equations and parameters used in the model are listed. We examined three categories of model publications, beginning from broad and going to narrow. First, using a combination of MeSH terms, we searched PubMed for computational physiology models broadly—this resulted in over 6,500 publications, of which only a fraction was relevant. Next, we searched more specifically for cardiovascular models in PubMed—which returned over 1,000 publications. Finally, examined diabetes model publications identified in a review article3 as Clinical Models—which was a list of 96 models. From each of the three categories, we randomly sampled and screened publications for inclusion. We analyzed the content of 50 full-text publications from each of categories for model code availability.

Results: All but one model publication examined had no model code available (one publication referred to an external link that is no longer available). Furthermore, most model publications in the general and cardiovascular categories only listed a subset of equations and parameters used for model simulation. More than a third of model publications in the diabetes category only list a subset of the model equations and parameters. In this third category, about 7% of the models were included in BioModels library, but all of these were added and curated after publication. Thus, even for publications with curated models, a scientist simply reviewing the literature would have no easy way of finding these model codes.

Conclusion: Despite the push towards reproducibility of computational models, the vast majority of model publications do not provide sufficient information to reproduce the model simulations they describe. At a minimum, modelers and authors should indicate that source code is available along with some information about the language used. Better, modelers should include sufficient information in their publications so that the models can be reproduced and simulated by others; ideally, this would include submitting their code into repositories such as BioModels Database or the PMR. Moving forward, one approach to ameliorate this situation would be if journals and publishers require or at least encourage authors to submit model source code and all of the equations and simulation parameter as part of the supplemental material for the publication.

1. Yu T, Lloyd CM, Nickerson DP, Cooling MT, Miller AK, Garny A, et al. The Physiome Model Repository 2. Bioinformatics. 2011 Mar 1;27(5):743–4.

2. Li C, Donizelli M, Rodriguez N, Dharuri H, Endler L, Chelliah V, et al. BioModels Database: An enhanced, curated and annotated resource for published quantitative kinetic models. BMC Syst Biol. 2010 Jun;4:92–92.

3. Ajmera I, Swat M, Laibe C, Le Novère N, Chelliah V. The impact of mathematical modeling on the understanding of diabetes and related complications. CPT Pharmacomet Syst Pharmacol. 2013 Jul 10;2:e54.

First, differences in the sequences of the two strains were identified by performing SNP analysis. These differences were then used to find metabolic alterations affecting the virulence. By using FBA and in-silico gene knock-out, three genes were identified that could be responsible for the difference in pathogenicity between the laboratory and the highly pathogenic patient strain. These genes affect metabolic reactions that are associated with virulence in P. aeruginosa. Gene knock-outs of these three genes are momentarily being performed in a laboratory experiment to verify their metabolic relevance in virulence.

However, only a fraction of the genes of P. aeruginosa is included in the published model. Many additional genes are associated with virulence, differ between the sequenced laboratory and patient strains, or both. Therefore, the model is being extended to increase its predictive value.

References:

[1] European Centre for Disease Prevention and Control. Antimicrobial resistance surveillance in Europe 2015. Annual Report of the European Antimicrobial Resistance Surveillance Network (EARS-Net). Stockholm: ECDC; 2017. https://ecdc.europa.eu/en/publications-data/antimicrobial-resistance-surveillance-europe-2015#no-link

[2] Bartell JA, Blazier AS, Yen P, Thøgersen JC, Jelsbak L, Goldberg JB, and Papin JA. Reconstruction of the metabolic network of Pseudomonsas aeruginosato interrogate virulence factor synthesis, Nature Communications (2017). doi:10.1038/ncomms14631

The semantic annotation and modelling platform we have developed is a new contribution enabling scientists and clinicians to discover relevant models in the PMR and reuse them in other computational modelling initiatives and translational projects linking to real-world health data. We believe that this approach demonstrates how semantic web technologies and methodologies can contribute to biomedical and clinical research. Furthermore, novice modellers could use this platform as a learning tool. The source code is available on GitHub: https://github.com/dewancse/epithelial-modelling-platform

Results: Here, we present a first manually curated draft reconstruction of the metabolic network in Treponema pallidum ssp. pallidum towards a genome-scale metabolic model (GEM). At this time, the model iSM161 comprises 161 genes of 1,039 predicted open reading frames that are responsible for 239 reactions of 277 metabolites. The model is still under development and steadily updated. For the reconstruction, COBRApy has been used, where subsystem information is added and parsed as SBML groups extension using libSBML.

Discussion: Using this reconstruction, together with the application of COBRA methods, we anticipate to gather new insights into the pathogen’s physiology and pathology, and in how this spirochete, which has earned the designation of “stealth pathogen,” succeeds in making a living and eluding human’s immune defenses as well as cultivation attempts. It is planned to make the model available to the community in SBML format.

TRONCO is an R package for the inference of the evolutionary history of cancer mutations, which was developed by the BIMIB group at University of Milan-Bicocca. It includes different functions, which guide the user through the process of creating an analysis starting from mutation data. The work presented here concerns the development of cyTRON/JS, a web application which serves as a bridge between JavaScript and R TRONCO functions. It was conceived for two main purposes:

- Providing an interactive visualization of TRONCO models: while TRONCO graph display is static and cannot be modified, cyTRON/JS provides an interactive view of graphs, making it possible to directly retrieve information about genes involved in the study, by accessing widely-used public genome databases, and about the algorithms which led to the model reconstruction. The visualization is based on Cytoscape.js. - Making TRONCO accessible to users not familiar with R programming. The application is made up of two main modules: one for data input and parameters setting and the other for model reconstruction, which is completely transparent to users. The communication between the modules is based on standard data exchange protocols and data formats.

cyTRON/JS was developed within the Google Summer of Code 2018 program, in collaboration with the National Resource for Network Biology (NRNB) organization.

Our online platform permits the attachment of sources of evidence, including data-sets, statistical calculations, images, and text, to model file components. Each attachment is accompanied by a value indicating confidence in that evidence, retaining the reasoning for model composition while identifying potential areas of deficiency.

Our tool is compatible with a range of model specification formats, including MATLAB, XML, and SBML. For SBML models, our platform permits location and attachment of resources in the Minimum Information Required in the Annotation of Models (MIRIAM) registry, using services provided by Identifiers.org

Overviews of evidence coverage and summary statistics are produced to assist in determining overall model confidence. Platform adoption offers benefit in increasing model quality, auditability, and confidence in wider model application, while reducing model maintenance effort and ensuring retention of IP within institutions.

For this reason, there is a need to clean data towards tasks such as disease modeling. This work will discuss a set of Python tools that were used to process ClinicalTrials.Gov and aim towards standardization. The tools parse XML data, index the information for easier processing, and then uses Machine Learning and Natural Language Processing to cluster the data, The clustering algorithm is used to assist a human user look at similar information provided by a Graphical User Interface.

These Python tools were used to extract 21,094 units from 30,763 different clinical trails with results. These quantities show a clear need of standardization to be used in future computer modeling efforts.

Here, we present an XML extension to Simulation Experiment Description Markup Language (SED-ML) that enables description of simultaneous fitting of several models and data sets, and describing uncertainty in the fitting. An extended SED-ML file describes several models that are simultaneously fitted to one or more data sets. Several new elements are added to SED-ML schema: (a) Cross-reference between models and data; (b) Several new tasks (such as optimization and ensemble simulations), with multiple layers added within each task, such as the parameters layer under optimization task, or histograms and subplots under ensemble simulations; (c) saving the set of the system, necessary to avoid extensive repeat of optimization if only post-processing is required. The simplified version of the extension is implemented as an input for the SloppyCell software (Gutenkunst, R.N., et al. 2007, PLoS Comp Biol) for exploring uncertainties in both model parameters and model predictions.