CNN based model for enzyme-substrate activity prediction using enzyme sequence and substrate structural information. EnzRank is used for rank-ordering novel enzyme-substrate activities identified in de novo biosynthesis pathway design and enzyme re-engineering projects.
User interface is available at "https://huggingface.co/spaces/vuu10/EnzRank"
- Tensorflow 2
- rdkit
- pandas
- tqdm
- numpy
- keras
- scikit-learn
- streamlit
- create a conda environment using:
conda create --prefix MLenv - activate the created environment using:
conda activate MLenv - install packages using requirement.txt file:
pip install -r requirement.txt - install rdkit using:
pip install rdkit - install streamlit using:
pip install streamlit
Three data inputs are required- proteins sequence, substrate smiles, and enzyme-substrate pairs (Look at CNN_data or CNN_split_data folder for the required input files)- input data must follow the same format
For performance prediction on different splits of data (CNN_data_split folder) run.
python predict_with_model.py ./CNN_results/CNN_results_1/model.model -n predict -i ./CNN_data_split/CNN_data_1/test_dataset/test_act.csv -d ./CNN_data_split/CNN_data_1/test_dataset/test_compound.csv -t ./CNN_data_split/CNN_data_1/test_dataset/test_protein.csv -v Convolution -l 2500 -V morgan_fp_r2 -L 2048 -W -o ./CNN_results/CNN_results_1/output_predictions.csv
-- Please use "predict_with_model_less.py" if you have less than 50 enzyme-substrate
This will output "output_prediction.csv" file in the CNN_results folder
To launch the streamlit based graphical user interface:
move to 'Streamlit' directory using- cd Streamlit
launch interface using command: streamlit run main.py
usage: EnzRank.py [-h] [--test-name [TEST_NAME [TEST_NAME ...]]]
[--test-act-dir [TEST_act_DIR [TEST_act_DIR ...]]]
[--test-mol-dir [TEST_mol_DIR [TEST_mol_DIR ...]]]
[--test-enz-dir [TEST_enz_DIR [TEST_enz_DIR ...]]]
[--with-label WITH_LABEL]
[--window-sizes [WINDOW_SIZES [WINDOW_SIZES ...]]]
[--enz-layers [enz_LAYERS [enz_LAYERS ...]]]
[--mol-layers [mol_LAYERS [mol_LAYERS ...]]]
[--fc-layers [FC_LAYERS [FC_LAYERS ...]]]
[--learning-rate LEARNING_RATE] [--n-epoch N_EPOCH]
[--prot-vec PROT_VEC] [--prot-len PROT_LEN]
[--mol-vec mol_VEC] [--mol-len mol_LEN]
[--activation ACTIVATION] [--dropout DROPOUT]
[--n-filters N_FILTERS] [--batch-size BATCH_SIZE]
[--decay DECAY] [--validation] [--predict]
[--save-model SAVE_MODEL] [--output OUTPUT]
act_dir mol_dir enz_dir
act_dir Training act information [mol, target, label]
mol_dir Training mol information [mol, SMILES,[feature_name,..]]
enz_dir Training enz information [enz, seq]
--validation Excute validation with independent data, will give AUC and AUPR (No prediction result)
--predict Predict interactions of independent test set
EnzRank script has two mode, validation and predict mode.
In validation mode, performances (AUC, AUPR, threshold for AUC and AUPR) on each step and selected hyperparameters are recorded.
In test step, prediction results for given test dataset will be reported after training.
--test-name [TEST_NAME [TEST_NAME ...]], -n [TEST_NAME [TEST_NAME ...]] Name of test data sets
--test-act-dir [TEST_act_DIR [TEST_act_DIR ...]], -i [TEST_act_DIR [TEST_act_DIR ...]] Test act [mol, target, [label]]
--test-mol-dir [TEST_mol_DIR [TEST_mol_DIR ...]], -d [TEST_mol_DIR [TEST_mol_DIR ...]] Test mol information [mol, SMILES,[feature_name,..]]
--test-enz-dir [TEST_enz_DIR [TEST_enz_DIR ...]], -t [TEST_enz_DIR [TEST_enz_DIR ...]] Test enz information [enz, seq]
--with-label WITH_LABEL, -W WITH_LABEL Existence of label information in test act
You can input multiple datasets for validation or test with argument specifier.
In addition to act information file, mol information file and target enz information file, you need to name of validation or test datasets.
For test dataset, you can inform that test dataset has label or not with -W value
--n-epoch N_EPOCH, -e N_EPOCH The number of epochs for training or validation
--prot-vec PROT_VEC, -v Convolution Convolution, can change this to use difference protein feature instead of convolution
--prot-len PROT_LEN, -l PROT_LEN enz vector length
--mol-vec mol_VEC, -V mol_VEC Type of mol feature (morgan_fp_r2)
--mol-len mol_LEN, -L mol_LEN mol vector length
--window-sizes [WINDOW_SIZES [WINDOW_SIZES ...]], -w [WINDOW_SIZES [WINDOW_SIZES ...]] Window sizes for model (only works for Convolution)
--enz-layers [enz_LAYERS [enz_LAYERS ...]], -p [enz_LAYERS [enz_LAYERS ...]] Dense layers for enz
--mol-layers [mol_LAYERS [mol_LAYERS ...]], -c [mol_LAYERS [mol_LAYERS ...]] Dense layers for mols
--fc-layers [FC_LAYERS [FC_LAYERS ...]], -f [FC_LAYERS [FC_LAYERS ...]] Dense layers for concatenated layers of mol and enzyme layer
--n-filters N_FILTERS, -F N_FILTERS Number of filters for convolution layer, only works for Convolution
--activation ACTIVATION, -a ACTIVATION Activation function of model
--dropout DROPOUT, -D DROPOUT Dropout ratio
--batch-size BATCH_SIZE, -b BATCH_SIZE Batch size
--learning-rate LEARNING_RATE, -r LEARNING_RATE Learning late for training
--decay DECAY, -y DECAY Learning rate decay
--save-model SAVE_MODEL, -m SAVE_MODEL save model
--output OUTPUT, -o OUTPUT Prediction output
For an example, if you have training dataset, CNN_data_split/split_0/training_dataset/training_act.csv, CNN_data_split/split_0/training_dataset/training_compound.csv and CNN_data_split/split_0/training_dataset/training_enz.csv with right specification.
You can validate model with validation dataset, CNN_data_split/split_0/validation_dataset/validation_act.csv, CNN_data_split/split_0/validation_dataset/validation_compound.csv and CNN_data_split/split_0/validation_dataset/validation_enz.csv by using this command line.
python EnzRank.py ./CNN_data_split/split_0/training_dataset/training_act.csv ./CNN_data_split/split_0/training_dataset/training_compound.csv ./CNN_data_split/split_0/training_dataset/training_enz.csv --validation -n validation_dataset -i ./CNN_data_split/split_0/validation_dataset/validation_act.csv -d ./CNN_data_split/split_0/validation_dataset/validation_compound.csv -t ./CNN_data_split/split_0/validation_dataset/validation_enz.csv -W -c 512 128 -w 10 15 20 25 30 -p 128 -f 128 -r 0.0001 -n 30 -v Convolution -l 2500 -V morgan_fp_r2 -L 2048 -D 0 -a elu -F 128 -b 32 -y 0.0001 -o ./validation_output.csv -m ./model.model -e 1
This command will train model with given hyper-parameters for 1 epoch. (because of -e 1).
And resulting in validation result ./validation_output.csv, and corresponding model ./model.model
To Look for the performance results of the negative test datasets generated using highly similar substrates-
on terminal run, "chmod +x test_predict_split_new.sh" then, run "./test_predict_split_new.sh"
To look for the performance evaluation and statistical analysis done to test if positive datasets ranks higher than newly generated challenging dataset-
Go to "CNN_results_split_final/test_dataset_new" folder and look for "Evaluate_prediction_10_per.ipynb" jupyter notebook which has all the necessary information
Please check the "bulk_retrieval.ipynb" jupyter notebook