BioModels: Content, Features, Functionality, and Use

Abstract

BioModels is a reference repository hosting mathematical models that describe the dynamic interactions of biological components at various scales. The resource provides access to over 1,200 models described in literature and over 140,000 models automatically generated from pathway resources. Most model components are cross-linked with external resources to facilitate interoperability. A large proportion of models are manually curated to ensure reproducibility of simulation results. This tutorial presents BioModels' content, features, functionality, and usage.

Figure 1

Source of literature models. As of September 2014, 1,212 models originate from 200 scientific journals. The inner layer of the chart gives the percentage of models that originate from journals listed in the second layer. For example, from the blue segment of the chart it should be understood that 27% of the models come from three journals (i.e., more than 100 models from each of the three journals). Approximately 50% (611 models) of the models are from 10 journals, of which 6 journals recommend model submission to BioModels in their authors' instructions (indicated by stars). This diagram not only illustrates the diversity of journals from which models are obtained, but also the importance of journal's support to the resource through encouraging authors to submit their models to BioModels, during the paper submission process.

Figure 2

Model production pipeline of BioModels. (A) The diagram schematically represents the model generation pipeline of the two branches ((a) models published in literature and (b) Path2Models) of BioModels. The literature models undergo a sequence of steps from submission until publication in BioModels. Models imported from pathway resources are converted to SBML with additional information (see B for details), and are submitted to BioModels. (B) Detailed representation of the workflow of models generated from different pathway resources. Mathematical features, such as kinetic rate equations and flux bounds are added during the process, along with a graphical description.

Figure 3

GO classification of models. All models are annotated with at least one GO term at the model level to represent the biological process it describes. In order to provide a broader level of classification of models based on these precise GO terms, three increasingly generic ancestor terms were chosen, the third term being one of the GO root term “Biological Process” (GO:0008150), which applies to all models. This type of hierarchical classification (from generic to most specific terms, through four GO terms) allows users to narrow down the search to specific biological process of interest. The top three GO terms used for the classification are shown, whereas the fourth term is the specific GO term used in the models. This allows categorized searching of models by following the GO term classification chart, and this facility is available through the BioModels website.

Figure 4

Importance of sharing and re-use of models. Clusters of models created based on the qualifier “bqmodel:isDerivedFrom” (see text for more information). The largest clusters are formed by models of signaling pathways. This diagram shows how signaling models are shared and re-used by different groups to develop more comprehensive models. The network is formed by the cross-talks of a range of signaling pathways activated downstream of ErbB, FGFR, NGFR, Insulin receptor, NMDAR, and AMPAR. The arrows (→) are used to link models that are derived/adopted from one or more parent models. Models available from BioModels are highlighted in salmon, otherwise in green. The models are named with the first author's surname followed by the year of publication. For some models, Author'sLastNameYear is followed by a number in brackets denoting the number of models in BioModels coming from the same publication. For example, “Bidkhori2012(2)” means that two variants of the model are described in the paper and both are available in BioModels. The two models “Bidkhori2012 - normal EGFR signalling” (BIOMD0000000452) and “Bidkhori2012 - EGFR signaling in NSCLC” (Bidkhori et al., 2012, BIOMD0000000453), describe the cross-talk mechanism between Ras/ERK, PI3K/AKT, and JAK/STAT pathways in normal and NSCLC non–small-cell lung cancer (NSCLC) cells, respectively. As can be seen from the figure, the model is itself derived from several other models.

Figure 5

Screen image of a “Model” page. This displays the landing page of “Proctor2013 – Effect of Aβ immunization in Alzheimer's disease” (BIOMD0000000488). The page displays model download formats (Download SBML, Other formats (autogenerated)), links to features such as Online simulations (Actions), distribution of model components that span various tabs (Model, Overview, Math, Physical entities, Parameters, Curation), and model elements cross-linked (annotated) with various external resources such as Gene Ontology terms and Human Disease Ontology.

Figure 6

Evolution of BIOMD0000000488: Amyloid-beta (Aβ) immunization (Proctor et al., 2013). Under stress or age-related conditions, a correctly folded native protein (NatP) can become misfolded (MisP). There are three possible outcomes for misfolded proteins; the protein may be refolded into its native state with the help of chaperones such as HSP90, may be degraded and removed from the system, or may aggregate together. Such aggregates are the main characteristics of most neurodegenerative diseases, such as Alzheimer's disease and Parkinson's disease. The models in this figure describe to varying extents these three mechanisms. The arrows (→) are used to link models that are derived/adopted from one or more parent models. The models are named as described previously (Author'sLastNameYear) and numbers in parentheses denote the number of models in BioModels which result from the same publication. The mechanism described in each model is recorded adjacent to the model name along with its BioModels Identifier. The shaded portion in the figure depicts the model lineage for “Proctor2013” (BIOMD0000000488), which was built using components of previous models. Models in the unshaded regions do not contribute to the lineage of Proctor2013, but may be relevant for the lineage of other models. See text for further information. GSK3, glycogon synthase kinase 3; HSP, heat shock protein; ROS, reactive oxygen species.