ceferisbarov / deepcas13 Goto Github PK

DeepCas13, A deep learning model to predict the CRISPR-Cas13d sgRNA on-target activity with high accuracy from sgRNA sequences and RNA secondary structures.

DeepCas13 uses convolutional recurrent neural network (CRNN) to extract spatial-temporal features for both sequence and secondary structure of a specific sgRNA and then predicts the efficiency by a fully connected neural network (FCNN).

We also provide a webserver for user to design Cas13d sgRNAs online.

1. Create a conda environment

	conda create --name deepcas13 python=3.6

2. Activate the environment

	conda activate deepcas13

3. Install the dependencies

    conda install pandas numpy=1.19.1 seaborn scikit-learn 
	
	conda install -c bioconda viennarna
	
	pip install jupyterlab
	
	pip install tensorflow==2.4
	
	

Here, we provie 3 demos to show how to use DeepCas13, including train model, predict sgRNA efficiency and design sgRNAs for target sequence.

We can train DeepCas13 model by following command:

	python deepcas13.py --train --savepath DL_model --data data/training_data.csv

The input training data is a two-column file, in which the first col is sgRNA sequence and the second col the LFC.

If the sgRNAs already exist, we can load pretrained model and run DeepCas13 to predict the efficiency:

	python deepcas13.py --seq data/test_data_sgrna1.csv --model DL_model

To design sgRNA for a specific target sequence, DeepCas13 can identify all possible sgRNAs and predict the efficiency:

	python deepcas13.py --seq data/test_data_target.fa --model DL_model --type target --length 22

The output of DeepCas13 is a csv file:

• Xiaolong Cheng: [email protected]
• Wei Li: [email protected]

Licensed under the MIT license. This project may not be copied, modified, or distributed except according to those terms.