Optimized sgRNA design by deep learning to balance the off-target effects and on-target activity of CRISPR/Cas9

The CRISPR/Cas9 system derived from bacteria especially Streptococcus pyogenes (SpyCas9) is currently considered as the most advanced tool used for numerous areas of biological study in which it is useful to target or modify specific DNA sequences. However, low on-target cleavage efficiency and off-target effects impede its wide application. Several different sgRNA design tools for SpyCas9 by using various algorithms have been developed, including linear regression model, support vector machine (SVM) model and convolutional neuron network model. While the deep insight into the sgRNA features contributing for both on-target activity and off-target still remains to be determined. Here, with public large-scale CRISPR screen data, we evaluated contribution of different features influence sgRNA activity and off-target effects, and developed models for sgRNA off-target evaluation and on-target activity prediction. In addition, we combined both activity and off-target prediction models and packaged them as an online sgRNA design tool, OPT-sgRNA.