Fig. 7From: Deep learning-based transcriptome data classification for drug-target interaction predictionThe predictive errors between other models and the DNN. Each dot represents the difference between the DNN and other models. If the dot had a negative label, the predictive error was from the CM i,0 of other models minus the CM i,0 of the DNN. Otherwise, the predictive error was from the CM i,1 of other models minus the CM i,1 of the DNN. Therefore, if more dots were below the horizontal line, the performance of the DNN was better than the performance of other models. Otherwise, the DNN performance was worseBack to article page