-
(1) [Python 3.8.x] Follow the instructions at: https://www.python.org/downloads/
-
(2) [Jupyter] Follow the instructions at: https://jupyter.org/install
-
(3) [scikit-learn 0.23] Run this command via PIP to install the proper version:
pip install scikit-learnFurther details at: https://scikit-learn.org/stable/index.html
-
(4) [pandas 0.25] Run this command via PIP to install the proper version:
pip install pandasFurther details at: https://pandas.pydata.org/
-
(5) [tqdm] Run this command via PIP to install the proper version:
pip install tqdmFurther details at: https://github.com/tqdm/tqdm
Please launch Jupyter from the terminal with the following command:
Jupyter notebookNow, you can simply execute each cell in the methods.ipynb notebook to reproduce the results shown in the following paper [On the use of topological features of metabolic networks for the classification of cancer samples. Machicao, J.,Craighero, F., Maspero, D., Angaroni, F., Damiani, C., Graudenzi, A., Antoniotti, M. & Bruno, O. M.].
The notebook populates a pandas dataframe with all the performance metrics used to perform the hyperparameters selection and to compare the classification methods.
We considered 105 breast cancer patients with gene expression profile obtained from either cancer and normal/health tissue, for a total of 210 gene expression profiles (source TCGA-BRCA database).
In the breast_cancer_expression directory one can find those data in two compressed files, one includes cancer and one includes normal samples.
To reproduce the classification analysis we provide in data directory also a file named R22_breast_cancer.csv in which each row is a sample and the columns include:
-
(1) The first 5 PCs computed considering the expression profiles of the metabolic genes included in Recon2.2 [https://doi.org/10.1007/s11306-016-1051-4].
-
(2) The extracted topological properties from each network pruned with a given threshold.
-
(3) The cancer / normal ground truth sample labels