Table 3. Internal Validation Metrics Reported as TPR % and TNR % for Individual Systems in the Bound/Unbound States^a.

		original training set		extended training set
ML models	TRAP1 complexes	TPR % (state I)	TNR % (state A)	TPR % (state I)	TNR % (state A)
GNB	5	71.4		49.8
	7	63.0		52.0
	6	61.2		62.1
	inhibitor-free (rep. 1–3)		69.9		75.1
	inhibitor-free (rep.4–6)		84.5		80.3
	inhibitor-free (rep.7–9)		90.0		80.2
KNB	5	70.9		78.3
	7	94.2		63.9
	6	66.6		68.8
	inhibitor-free (rep. 1–3)		70.1		69.5
	inhibitor-free (rep.4–6)		89.6		83.4
	inhibitor-free (rep.7–9)		82.9		78.4
GDF-SVM	5	92.5		90.9
	7	97.3		93.7
	6	95.4		91.3
	inhibitor-free (rep. 1–3)		95.7		92.7
	inhibitor-free (rep.4–6)		98.4		96.7
	inhibitor-free (rep.7–9)		92.5		89.7

^aPercentages are shown over chunks of 12 000 (original training set) or 18 000 (extended training set) frames. Models trained on the entire training set were used to make predictions.