The code developed for the preprint manuscript: A systematic evaluation of data processing and problem formulation of CRISPR off-target site prediction
The code contains multiple options of training and testing models, including a transfer learning model for CHANGE-seq to GUIDE-seq.
Once you have cloned this repository and installed all the requirements, unzip the potential off-target sites dataset located in:
files/datasets/output_file_pam_change.zip
Then, you should execute
python prepare_data.py
to generate the active and inactive datasets that will be used in the training and testing stages.
To train the models, you should execute
python main_train.py
This will train some examples models. You can modify the main_train.py to run all the possible options we have formed.
Once you trained your models. you can evaluate the prefoamce of the models by executing
python main_test.py
In the current form, main_test.py will evaluate the example models and save the prediction performance tables located in files. To evaluate other models variants, please modidy main_test.py accordingly.
The code was tested with:
Python interpreter == 3.6.6
Python packages required (other versions may work as well):
numpy == 1.18.5
pandas == 1.1.2
biopython==1.78
scikit-learn == 0.23.0
xgboost==1.3.3
Note: There might be other packages needed. Please contact us in case of any problem.