Fig. 3.

Comparing the performance of the proposed deep learning model with existing models and a supervised model. a ROC curve in GM12878 dataset. b ROC curve in HepG2 dataset. c ROC curve in K562 dataset. d AUC values showing the performance of the proposed model and a supervised model without pseudo labels in K562 dataset. AUC, area under the ROC (receiver operating characteristic) curve; Supervised_dnn, deep neural network; SSL_dnn, semi-supervised learning by dnn with pseudo labels; $\mathrm{\alpha }(\mathrm{t})$, a parameter in the loss function