Fig. 3.

Performance evaluation of essential gene classification of training sets (self-predictions) within selected eukaryotic species using Area Under Receiver Operating Characteristic and Precision-Recall Curves (ROC-AUC and PR-AUC; training set sizes between 10 and 90%, using 10% increments). The dots represent the calculated ROC-AUC/PR-AUC values, and linear models fit dots representing the performances of each machine-learning algorithm. Feature selection procedures were performed for each subsample.