The outputs of some scripts are liable to change due to f.ex. changed content in databases from when they were last run. Files for the intermediate steps are therefore included.
For some steps, external dependency versions can be relevant. We therefore include the environment folder that contains lists over all installed dependencies.
Python 3.10 was installed inside a conda environment, while dependencies inside the environment were installed via pip.
Once the general environment is set up, install the packages inside code_packages. This can be done from the packages' folder using pip install . .
Of the two packages provided, cofactors provides the relevant modeling code, while plotting only contains helper functions for visualization.
The proteomics data from mitoPARP cell lines and parental HEK293 used are included under external_data,
the raw data contained in proteomics.xlsx, the processed data integrated into the models are in 293parp_abundance_ratios_mapped.csv.
uniprot_acc_to_ensemble_gene.tsv contains mappings to convert between the identifiers in the data and the model.
They were downloaded from taken from david.ncifcrf.gov.
All the steps performed can be followed in code_modeling/extract_proteomics.ipynb
Kms can be downloaded to a default location using the script ./code_modeling/download_kms.py.
This script relies on having the cofactors package installed.
Note that the database contents may change over time as more data is added.
external_data/mitocore contains the Mitocore model used, along with metadata and data extracted from the models.
See also the official source.
Building the models happens both inside Matlab (proteomics integration) and Python (NAD integration).
The process therefore involves multiple scripts.
There are also models created for the supplemental data.
Relevant code is in code_modeling.
code_modeling/integration_gimme.m creates the 293mitoPARP and parental HEK293 models that are used in the main analyses.
code_modeling/paramscan_gimme.m creates the necessary series of models to estimate the effects of different objective fractions on the GIMME models.
code_modeling/integration_nad.ipynb creates the models parameterized for NAD levels.
Here, we also solve the models with FBA and pFBA and save the results to be used later.
Running this script relies on the cofactors package.
Some of the analyses (solving the models with pFBA and FBA) is already run along with the model creation.
The scripts here contain the concrete analyses performed with the solutions along with the visualizations.
The relevant scripts are contained in code_analysis.
The core analyses are performed in analyses_parp.ipynb, where the illustrations related to the 293mitoPARP / HEK293 models are also created.
Running the analyses depends on having the plotting package installed.
check_kms.ipynb gives an overview over the Kms extracted from Brenda and SabioRK.
comparison_brenda_sabiork.ipynb shows how using the Brenda or SabioRK databases for obtaining Kms influences the models.
gene_mapping.ipynb analyzes how many of the IDs in the proteomics dataset could be mapped to the model and vice versa.
graphing_gimme_mapping.ipynb gives an overview of how many of the reactions in Mitocore were ultimately covered by the proteomics integration.
parameter_scan_c0.ipynb tracks model results across a range of different C_0 parameters.
scan_params_gimme.ipynb tracks model results across a range of GIMME parameters.
scan_decision_functions.ipynb compares the results of different decision functions when choosing between multiple Km values.
scan_fva_frac.ipynb tracks model results across a range of FVA objective fractions.
The results folder contains the direct analysis results from the models created.
In the folder itself the data are sorted by metric, as they are used for the analyses in the main text.
results/supplements contains similar data, sorted by model instead to give a better overview of the calculations.