Table 1.

Brief network framework description of several pre-selected models

Model_Name	Description
CnnCrispr	The final model including all units we mentioned
CnnCrispr_NoLSTM	Without biLSTM layer
CnnCrispr_Conv_LSTM	Reverse the order of the convolution layer and the recurrent layer
CnnCrispr_NoBatchNor	Without Batch Normalization layer
CnnCrispr_NoDropout	Without Dropout layer

CnnCrispr is the benchmark model. On this basis, some parts were removed to obtain the comparison models. With the exception of parts mentioned in the description, the structure of the contrast model was completely consistent with the benchmark model. The CnnCrispr_Conv_LSTM model replaced the sequence of convolutional layers and recurrent layer to compare the influence of network sequence on the prediction results