Fig. 1: Performance metrics for predicting low-coordination CH and EDH sites. | Nature Communications

Fig. 1: Performance metrics for predicting low-coordination CH and EDH sites.

From: PinMyMetal: a hybrid learning system to accurately model transition metal binding sites in macromolecules

Fig. 1

Performance metrics for predicting low-coordination CH (a–c) and EDH (d–f) sites. a, d Receiver operating characteristic (ROC) curves show strong model performance, with AUC values of 0.99 for CH and 0.97 for EDH. b, e Precision-recall (P-R) curves highlight the ensemble model’s superior balance, achieving 0.98 and 0.94 average precision for CH and EDH, respectively. c, f Confusion matrices show higher accuracy for CH sites, with EDH predictions proving more challenging.

Back to article page