Overview

This project is intended to use deep learning models for Crispr-Cas on-target efficiency prediction and off target specificity prediction.

Below is the layout of the whole model.

AttnToMismatch_CNN

This model includes four components:

• Embedding layer
• Transformer layer
• Convolutional neural network
• Fully connected layer

AttnToCrispr_CNN

This model includes four components:

• Embedding layer
• Transformer layer
• Convolutional neural network
• Fully connected layer

Requirement

1. if conda is used, the virtual environment can be created with:

conda env create -f environment.yml

2. required packages

• keras
• tensorflow
• pytorch
• sklearn
• pandas
• numpy
• skorch
• visdom
• shap

Usage

Specify which data or model to use, such as cpf1 and cpf1_OT.

python ./attn_to_crispr.py <data/model>

<data/model> could be K562/A549/NB4/cpf1/cpf1_OT/deepCrispr_OT

Training new model with customized dataset

a. Off target prediction on customized dataset

1. Organize dataset format as the example dataset in dataset/customized_Cas9_OT
2. Save the new dataset as dataset/customized_Cas9_OT/customized_Cas9_OT_data.csv

python flexible_OT_crispr.py customized_Cas9_OT

4. Optional: Specify training-testing split methods: change split_method in "models/customized_Cas9_OT/config.py":

• "regular" for n-fold split
• "stratified" for leave sgRNAs out split

b. Update: support on target prediction on customized dataset without only sgRNA sequence features

1. Organize dataset format as the example dataset in dataset/customized_Cas9_ontar
2. Save the new dataset as dataset/customized_Cas9_OT/customized_Cas9_ontar_data.csv
3. make sure the extra_numerical_features variable in "models/customized_Cas9_ontar/config.py" file is extra_numerical_features = [], this indicates no extra features are added besides sgRNA sequence features

python crispr_attn.py customized_Cas9_ontar