PropertyValue
is nif:broaderContext of
nif:broaderContext
is schema:hasPart of
schema:isPartOf
nif:isString
  • The domain model utilises the empirical findings of Carlotti et al [29, 30] and Yang et al [31, 32], with the statistical analysis being specifically based on a subset of data from Yang et al [32]. This subset contained measurements from single-cell analysis performed on 88 cells: 52 were transfected with IκBα Enhanced Green Fluorescent Protein (EGFP) and stimulated with IL-1; and 36 were transfected with IκBα-EGFP, but not stimulated with extracellular ligand, thus representing a control group. Single-cell analysis on live cells, include continuous monitoring of the same set of cells over time. All measurements within the data related to cytoplasmic fluorescence and were taken over a period of one hour, at intervals corresponding to 0, 10, 30 and 60 min. The subset of data used within the statistical analysis of this manuscript can be found in S1 and S2 Tables of the supplementary information. The data sets were divided into 3 groups based on transfection levels of the exogenous protein as in the original analysis by Yang et al: 0-1.5 fluorescent units (corresponding to up to 4 fold levels of the endogenous protein); 1.5-3.0 fluorescent units (4-8 fold levels of the endogenous protein); and above 3.0 fluorescent units (above 8 fold levels of the endogenous protein) [32]. The domain model was developed in an iterative manner by the modeller (RAW), senior software engineer (JT) and domain expert (EEQ), using the deep-curation approach [33]. We have chosen to follow the approach of Read et al [13] in using UML as the basis to semi-formally define the domain model of our complex dynamical biosystem. Along with a number of UML diagrammatic notations (UML v2.4, [10]), a number of less formal cartoon diagrams were also used to ensure the biological meaning could be conveyed efficiently. Furthermore, a number of statistical techniques were used to complement UML when modelling the temporal dynamics and stochastic characteristics of the system. Initial focus is the emergent system-wide behaviours of the pathway, before increasing the level of detail to the interactions between system components, and then the dynamics of individual components. The domain model is presented in a top-down manner, comprising three levels of abstraction, as defined below: A system-level overview of the domain model. This highly abstract level provides an outline of the biology of the IL-1 stimulated NF-κB signalling pathway. Particular focus is made to the behaviours of the system following induction by extracellular signal, and how these are believed to correspond to phenomena observed in the real-world domain. This abstraction level of the domain model does not make use of UML, but instead utilises less formal cartoon diagrams to convey system-wide properties, along with a number of statistical approaches to convey temporal dynamics. In particular, we have used the R data analysis and graphics software [34] to perform Chi-squared goodness of fit tests to ascertain the statistical distribution of the wet-lab data, along with hierarchical clustering and Principal Component Analysis to investigate any underlying groupings with the dataset of Yang et al [32] (see supplementary information for detailed descriptions).Modelling component-level interactions of the domain model. This medium level abstraction, decomposes the IL-1 stimulated NF-κB signalling pathway into its constituent molecular components. This level models an abstracted view of the key molecular interactions between the components, that together give rise to the emergent behaviours of the system. A cartoon diagram, along with UML communication and activity diagrams have been used in modelling these component-level interactions.Modelling individual component dynamics. This level of abstraction provides the greatest detail within the domain model, through modelling the dynamics of individual components within the system. A set of linked UML state machine diagrams have been used to develop this level of the model. Validation of the Domain Model: The iterative approach to developing the domain model within a CoSMoS project provides an ability to use a number of validation techniques during model formulation. Balci [35] has extensively reviewed the verification and validation techniques that are suitable for computational model development and simulation-based experiments. We have used a subset of these techniques to validate our domain model, comprising: audits by the senior software engineer to ensure that the modelling adheres to established practices; desk checking by the modeller to ensure that individual diagrammatic and statistical models are correct, complete, consistent and unambiguous; face validation by the domain expert to compare the complete domain model against her detailed understanding and judgment of the real-world biological system; and structured walkthroughs by the whole group (modeller, senior software engineer and domain expert) to detect and document faults.
rdf:type