Fig. 9

Comparison of the performance of our method with five earlier state-of-the-art methods on four mouse embryonic tissues. A ROC curves in the mouse tissues of our LR model trained on the CRM+TF⁺/Non-CRM sets in the mESCs or in the 67 human cell/tissue types. Notably, the performance of the two models is almost indistinguishable. B AUROCs in the mouse tissues of our LR models trained on the mESCs or on the human cell/tissue types in comparison with those of the five earlier methods. REPTIPLE, RFECS, DELTA, and CSI-ANN models were trained on the positive/negative sets (Table 2) in the mESCs as reported in reference [39], and the Matched Filter model was trained on the positive/negativize sets (Table 2) in the Drosophila S2 cells as reported in reference [62]. Experimental data reported in references [39, 62] were used for the predictions in the four mouse tissues