Bayesian uncertainty quantification

The German Network for Bioinformatics Infrastructure - de.NBI offers first class bioinformatics services including training and education to users in basic and applied life sciences research. In this network 40 projects belonging to eight service centers provide services that cover a wide variety of methods (genomics, proteomics, ...) and applications (from plants to humans). de.NBI-SysBio is the Systems Biology Service Center of de.NBI. In collaboration with FAIRDOM, de.NBI-SysBio serves the ...

The group around Nicole Radde specializes in the modeling, analysis, and simulation of biochemical systems. This especially includes parameter optimization and identification.

This code uses the PEtab problem (written in the yaml file) to perform MCMC sampling with pyPESTO. Afterwards an ensemble is created, that allows to compute the posterior predictive distribution and therefore the credibility intervals for the model output.

The sampling.py scipt of pyPESTO visulize was adjusted for an improved visualization. The used code was also uploaded.

A Conda environment file (.yml) containing the specific version of Python and installed packages with the according versions, ...

Complete posterior distributions for each drug and condition.

The files are in the hdf5 format and contain the complete information of the analysis for a FAIR data sharing.

The parameters A and B were estimated for the function f(x)=x⋅e^(B-Ax) and the AUC was calculated for the decay phase only. For OH-Midazolam the fourth replicate in the 2 weeks condition has an outlier in the measurement after six hours (more than 2SD above the mean of this condition) and was omitted for the AUC calculation as this could not be fitted to an exponential decay.

For each drug and condition, there is one:

• Visualization of the marginals and the sampling traces
• The credibility intervals for the parameters with the median
• Visualization of the individual ensemble output predictions with boxplots of the underlying data

Furthermore all conditions of the ensemble output predictions for one drug are visulized in one plot.

The exponential decay model with all parameters, observables and conditions was specified in a yaml file.

This yaml file is converted with yaml2sbml (2020 Jakob Vanhoefer, Marta R. A. Matos, Dilan Pathirana, Yannik Schaelte and Jan Hasenauer) to a PEtab problem, which contains also the SBML model.

The SOP creates a separate SBML model for each drug and condition, as the PEtab problem contains diffrent experimental data for them.

However, the SBML models only differ in their name as for all drugs and conditions, the same exponential decay model was assumed.

The SBMLs are automatically created by yaml2sbml, when the SOP is executed. Therefore, these files are for completeness only and are not necessary to replicate the analysis.