This repository contains two machine learning models and is populated with already trained models which can be used for immediate prediction with test data. Some details of the models are as follows:
- e3liagse_prediction.py predicts candidate E3 ligases for compounds.
- e3binder_prediction.py predicts if a compound is E3-binder or non-binder.
git clone https://github.com/Fraunhofer-ITMP/e3_models.git
cd e3_models
conda create --name=e3model python=3.9
conda activate e3model
pip install -r requirements.txt
To run the models, locate the src folder via terminal and type for instance 'python e3liagse_prediction.py'. The file to be predicted is located in 'input' folder of both models.
The models are trained with several ML-algorithms such as xgboost, random forest, naive bayes, linear regression, lightgbm and decision trees. The performance of each model is shown below.
The AUC-ROC curves of two best performing models are shown below.
