Probabilistic quotient's work & pharmacokinetics' contribution: countering size effect in metabolic time series measurements
Creation and comparison of PQN, PKM, and MIX models for size effect normalization.
For all details, read our research paper published at BMC Bioinformatics .
The size_effect_normalization
package requires Python 3.7 and packages listed in requirements.txt
. You can install the size_effect_normalization
package via setup.py
.
pip install -r requirements.txt
python setup.py install
To test if the installation was successful you can execute a (shortened) simulation run.
python tests/run_simulation_example.py
A tutorial for PKMminimal and MIXminimal normalization is available as Jupyter Notebook in docs/source/data/Tutorial.ipynb
.
A "Read the Docs" style documentation of the size_effect_normalization
package can be accessed here.
Scripts used to run simulatons v1-v3 (Figures 4-7, Supplementary Figures S8-S10) are are located in synthetic_data/run_simulation_v*.py
.
Results of the simulations are located in synthetic_data/simulation_results/*.pkl
.
An investigation on the performance of PQN on noisy data (Figure 8) is located in synthetic_data/Noisy_PQN.ipynb
.
Results of the simulation performed there is stored in synthetic_data/other_results/noisy_pqn.pkl
.
The script to test the difference of PQN and MIXminimal in performance on noisy data (Figure 9) is located in synthetic_data/comparison_with_noise.py
.
The results are located in synthetic_data/other_results/comparison_with_noise_results/*
.
The script to test the lambda hyperparameter (Supplementary Figure S3) is located in synthetic_data/search_lambda_v3.py
.
The results are located in synthetic_data/other_results/lambda_results/*
.
The script to test different loss function and transformation function combinations (Supplementary Figure S3) is located in synthetic_data/search_L_T_v3.py
.
The results are located in synthetic_data/other_results/L_T_results/*
.
An investigation of the influence on the weighting parameter lambda on the variance of fitted sweat volumes (Supplementary Figure S11) is done in synthetic_data/Lambda_Variance.ipynb
.
Results of the simulations are stored in synthetic_data/other_results/lambda_variance*.pkl
.
Jupyter Notebooks that run real data simulations are located at real_data/Brunmair_2021/PKM_Sub_2.ipynb
and real_data/Brunmair_2021/MIX_Sub_2.ipynb
for PKM and MIX respectively.
The results of these simuations are located in real_data/Brunmair_2021/PKM_sub_2/*
and real_data/Brunmair_2021/MIX_sub_2/*
for PKMminimal and MIXminimal respectively.
The script for data preprocessing is located in real_data/Brunmair_2021/Preprocessing.ipynb
.
The Jupyter Notebook that runs real blood plasma data simulations is located at real_data/Panitchpakdi_2021/DPH_Sub_2_models.ipynb
.
The results of these simulations are located in real_data/Panitchpakdi_2021/Sub_2_plasma.pkl
.
The script for data preprocessing is located in real_data/Panitchpakdi_2021/Preprocessing.ipynb
.
A Jupyter Notebook that replicates all figures used in the manuscript is located at /Figures.ipynb
.
All original code is licensed under the GNU GPL version 3.