Machine intelligence-driven framework for optimized hit selection in virtual screening

Table 3 CS-driven Stacked Ensemble framework and Benchmark performance comparison for CS module: the performance comparison of CS-driven Stacked Ensemble framework and Benchmark with individual ML algorithm

Algorithm	Dataset	Accuracy	Sensitivity	Specificity	AUC-ROC
Random Forest (RF)	Internal test set (x')^a	0.826	0.793	0.891	0.891
	Small independent validation dataset^b	0.726	0.642	0.747	0.807
	Large independent benchmark dataset^c	0.914	0.705	0.823	0.823
Extreme Gradient Boost (XGB)	Internal test set (x')^a	0.809	0.819	0.761	0.812
	Small independent validation dataset^b	0.789	0.571	0.816	0.782
	Large independent benchmark dataset^c	0.908	0.827	0.709	0.787
Deep Neural Networks (DNNs/DL)	Internal test set (x')^a	0.902	0.896	0.923	0.914
	Small independent validation dataset^b	0.894	0.877	0.782	0.866
	Large independent benchmark dataset^c	0.924	0.767	0.923	0.951
Stacked Ensemble	Internal test set (x')^a	0.948	0.961	0.988	0.991
	Small independent validation dataset^b	0.867	0.911	0.967	0.839
	Large independent benchmark dataset^c	0.962	0.921	0.987	0.902

^aTotal of 175 (81 inhibitors and 94 non-inhibitors) partitioned into 7:3 classified as Internal test set (x')
^b56 (43 inhibitors and 13 non-inhibitors) classified as small independent validation dataset
^c3415 (115 inhibitors and 3300 decoys (termed as non-inhibitors)) classified as large independent benchmark dataset

ISSN: 1758-2946