Table 4. Performance Metrics for the External Validation of the Three ML Algorithms Trained on the Original and Extended Dataset^a.

		GNB			KNB			GDF-SVM
training set	test set	TNR	TPR	r2	TNR	TPR	r2	TNR	TPR	r2
original	small	0.68	0.38	0.50	0.59	0.42	0.45	0.69	0.40	0.11
	large	0.67	0.41	0.64	0.58	0.44	0.61	0.68	0.42	0.53
extended	small	0.77	0.31	0.56	0.65	0.37	0.71	0.68	0.37	0.18

^aTNR (states A) and TPR (states I) are extracted from 96 000 and 102 000 MD frames for each state in the small and large test sets, respectively. The r² values obtained from linear regression analyses are also shown for each set of predictions, with the best values highlighted in red.