Fig. 5

Prediction of Cpf1 guide RNAs off-target specificities using deep convolutional neural networks. a ROC curves used to assess the performance of DeepCpf1 with fivefold external cross-validation. b ROC curves and AUC values comparing the performance of the CNN and other machine learning methods. c Comparison of the genome-wide specificities of AsCpf1 and SpCas9 nucleases. All possible sites with up to seven mismatches to 20 guide RNAs were identified by Cas-OFFinder tool. d The performance of various machine learning methods on independently generated Digenome-seq and GUIDE-seq data sets