Calibrating computational models of biological systems, assigning parameter values to ensure the model reflects behaviours observed biologically, can greatly impact the strength of hypotheses the model generates. A calibrated model provides baseline behaviour upon which sensitivity analysis techniques can be used to analyse potential pathways impacting model response. Where a behaviour depends on an intervention, model responses may be dependent on conditions at the point when the intervention is applied, complicating the calibration process and making it difficult to assess the extent to which an alteration in behaviour can be attributed to the intervention alone. Where a model is specified in Systems Biology Markup Language (SBML), there is a key deficiency in tools for solving models dependent on interventions.
ASPASIA, a cross-platform, open-source (GPLv2 license) Java toolkit, addresses this problem. ASPASIA can generate and modify models using SBML solver output as an initial parameter set, allowing interventions to be applied once a steady state has been reached. Additionally, multiple SBML models can be generated where a subset of parameter values are perturbed using local and global sensitivity analysis techniques, revealing the model's sensitivity to the intervention. As an exemplar of how ASPASIA can be used, we provide tutorials based on an SBML model of Th17-cell polarisation developed by YCIL PhD student Steph Dyson. A publication detailing the use of ASPASIA is currently in production.
Publication: Currently Under Review
To demonstrate the use of ASPASIA, we have developed a computational model to capture the dynamics of Th17 plasticity in vivo that can be used to infer the dynamics of a receptor and cytokine involved in the phenotype switching of Th17 cells. The model was developed in COPASI (Complex Pathway Simulator) and builds on work by Yates and Schulz to capture the dynamics of transcription factors T-bet and ROR-gamma in a CD4+ T cell undergoing polarisation and phenotype switching by interaction with exogenous cytokines. The model is in the process of being uploaded to the BioModels database.
(Full details of this model will be given in a paper that is currently in production)
A key novelty ASPASIA offers is to produce an SBML model from SBML solver output where that models behaviour has reached a steady state. This is done by reading in the final line of SBML solver output CSV file and setting the parameters or species concentrations to values in that file.
To introduce an intervention, a change in these parameter or concentration values is specified in the ASPASIA settings file. Simply specify the names of the parameters/concentrations that are being changed and the type of intervention being applied, with a value (replace, multiply by, subtract, add, divide by). You can do this manually in a text editor or using the Online ASPASIA Settings File Generator.
To generate SBML Model files containing the intervention:
A Robustness analysis examines each parameter or initial species concentration specified in the ASPASIA settings file in turn, perturbing the value assigned to that parameter or concentration within a specified range. The objective is to determine how robust the SBML model is to a perturbation in a single value, indicating the implications that biological uncertainty or parameter estimation may have when considering the impact of SBML model derived results.
To generate SBML Models for performing a Robustness Analysis using ASPASIA:
Local sensitivity analysis techniques cannot identify any compound effects where the influence of one SBML parameter or species concentration may rely on the value assigned to another. Such effects become evident by perturbing a number of parameter or species concentrations simultaneously, using global sensitivity analysis techniques. This approach is advantageous in revealing SBML parameters that have a strong influence on model behaviour.
ASPASIA can create SBML models for two global analysis techniques. The first is latin-hypercube model file generation, a sampling-based technique that selects values for the parameters or species concentrations specified in the ASPASIA setings file from a given value range, while aiming to minimise correlations in parameter values between chosen sets and efficiently covering the parameter space. The second is an implementation of the Extended Fourier Amplitude Sampling Test (eFAST), a technique that selects parameter values using sinusoidal curves of particular frequencies, with the aim to use that frequency to partition variance in model response between the parameters of interest. For both techniques, ASPASIA produces SBML model files for each parameter set, ensuring the sensitivity analysis can be performed on any SBML solver.
To generate SBML Models for performing Global Sensitivity Analysis using ASPASIA:
Note that for Latin-Hypercube analysis, you can provide a CSV file of parameter value sets if you wish. This is useful if you wanted to change a model yet run it with the same parameter value samples. To do this, see the description in the settings file.