Machine-Learning-Based Prediction of Chinese Hamster Ovary Cell Long-Term Stability Due to Epigenetic Changes
These are the datasets and model files associated with the publication Machine-Learning-Based Prediction of Chinese Hamster Ovary Cell Long-Term Stability Due to Epigenetic Changes. This work use ML models, primarily MLPs trained with the weighted focal differentiable MCC loss function, to predict the long-term stability of CHO cells based on the levels of chromatin/histone modifications over time. The models were trained on combined datasets generated by us.
The plots can be recreated by running the make_plots.py file with the appropriate flag as an input (python make_plots.py for Figs. 1, 4, and A1, python make_plots.py -c for Fig. 2, and python make_plots.py -b 1 0.5 0.333 0.25 0.20 0.10 for Fig. 3).
The Conda environment defining the specific packages and version numbers used in this work is available as environment.yaml. To use our trained model, create an Nx97 .csv file, without any headers, containing the levels of relevant chromatin/histone modifications over time, then run the Predict.py file as python Predict.py <path/to/file.csv> -t <threshold> -nc <number_of_classes> -bs <batch_size>. Results will be saved as a new .csv file.
If you have used any of the models in this work, please cite it. A Bibtex-formatted citation is available in citation.bib.