Fig. 2From: Deep learning improves the ability of sgRNA off-target propensity predictionVisualization of auPRC and Recall value in three test sets. CnnCrispr had higher auPRC and Recall values in the three test sets. Although the performance of CnnCrispr_NoDropout was similar to CnnCrispr, the addition of Dropout layer can reduce the network parameters, shorten the training duration and prevent over-fitting of the model to some extent, hence we finally chose the CnnCrispr with Dropout layerBack to article page