Fig. 1
From: Deep learning improves the ability of sgRNA off-target propensity prediction
The structure diagram of CnnCrispr. The other four pre-selected models were obtained by adding some parts on the basis of this frame. a Rules for setting index values for different r-d pair, and an example of index representation for a sgRNA-DNA sequence pair. b Based on index representation, the unsupervised GloVe model was used to train the embedded vectors and embed the sequence information into the new input matrix. c biLSTM layer was used to capture context information in input information. d CNN containing 5 layers with different kernels, scanning the above results to obtain different features. The last fully connected layers were used to obtain the final result