/Drug-Toxicity-Prediction-MultiLabel

Primary LanguagePython

Drug-Toxicity-Prediction-MultiLabel-Project

A multi-label learning model for predicting drug-induced toxicity in multi-organ based on toxicogenomics data.

Keywords

Drug-induced Toxicity, Multi-label, Prediction, Multi-organ, Gene Expression Data

Dataset

Toxygates, Open TG-GATEs, Pathology items

Requires

R version 3.4.0
Python version 3.6.6

keras==2.4.3
tensorflow==2.2.0
numpy>=1.17
scipy==1.4.1
Cython>=0.29
scikit-learn==0.23.2
setuptools==50.0.3
absl-py==0.10.0
matplotlib==3.2.0
joblib==0.16.0
h5py==2.10.0
pandas==1.1.2
six
liac-arff
arff
tqdm
astor

Installation

FILE: Drug-Toxicity-Prediction-MultiLabel/AttRethinkNet-Multilabel-Classification

Install packages

pip install numpy Cython

Compile and install the C-extensions Change Directory in CMD: change to project working directory using the "cd" command

python ./setup.py install
python ./setup.py build_ext -i

Run example in project directory

python ./examples/classification.py

Orgainzation

  1. AdvancedML-MLSMOTE Method : Multilabel Synthetic Minority Over-sampling Technique (MLSMOTE)

    • mlsmote-liver.py Produce synthetic instances for imbalanced multilabel liver dataset.
    • mlsmote-kidney.py Produce synthetic instances for imbalanced multilabel kidney dataset.

  2. Drug-feature-sorted-by-score-MLFS

    • LoadingData.py Provide basic operations to loading data.
    • MultiLabelFStatistic-liver.py Multi label F-Statistic (MLFS) algorithm in liver dataset.
    • MultiLabelFStatistic-kidney.py Multi label F-Statistic (MLFS) algorithm in kidney dataset.

  3. Drug-multilabel-classifier-comparison a. Liver dataset: Classifier : Binary relevance (BR), Classifier chains (CC).

    • drug_featureSelection_mlClassifier_lr_liver.py Base classifier: Logistic regression (lr).
    • drug_featureSelection_mlClassifier_rf_liver.py Base classifier: Random forest (rf).
    • drug_featureSelection_mlClassifier_svm_liver.py Base classifier: Linear support vector machines (svm).
    • integrative-model-jackknife-knn-liver.py Integrative model. b. Kidney dataset: Classifier : Binary relevance (BR), Classifier chains (CC).
    • drug_featureSelection_mlClassifier_lr_kidney.py Base classifier: Logistic regression (lr).
    • drug_featureSelection_mlClassifier_rf_kidney.py Base classifier: Random forest (rf).
    • drug_featureSelection_mlClassifier_svm_kidney.py Base classifier: Linear support vector machines (svm).
    • integrative-model-jackknife-knn-kidney.py Integrative model.

  4. AttRethinkNet-Multilabel-Classification

    • mlearn.models.rethinknet.rethinkNet.py (Att-RethinkNet framework)
    • examples (Run models and draw ROC curves, 002 means our proposed Att-RethinkNet, 001 means RethinkNet.)

How to Use

Traditional Methods: Download datasets->MLSMOTE->MLFS->Feature selection->Classification.
Att-RethinkNet Method: Download datasets->MLSMOTE->Att-RethinkNet Classification.