Large-scale comparison of machine learning methods for profiling prediction of kinase inhibitors

Wu, Jiangxia; Chen, Yihao; Wu, Jingxing; Zhao, Duancheng; Huang, Jindi; Lin, MuJie; Wang, Ling

doi:10.1186/s13321-023-00799-5

Journal of Cheminformatics

Table 2 Performance comparison results of the fingerprint-based models on the test sets of 354 kinases

From: Large-scale comparison of machine learning methods for profiling prediction of kinase inhibitors

Molecular feature	Method	AUC ^g	F1 score ^h	BA ⁱ
AtomPairs	RF^a	0.779 ± 0.161	0.736 ± 0.259	0.625 ± 0.124
	NB^b	0.733 ± 0.135	0.716 ± 0.186	0.680 ± 0.117
	SVM^c	0.698 ± 0.214	0.712 ± 0.286	0.620 ± 0.157
	KNN^d	0.743 ± 0.152	0.747 ± 0.222	0.665 ± 0.126
	XGBoost^e	0.759 ± 0.167	0.750 ± 0.212	0.653 ± 0.127
	DNN^f	0.752 ± 0.171	0.714 ± 0.238	0.631 ± 0.128
	Mean	0.744 ± 0.027	0.729 ± 0.017	0.646 ± 0.024
FP2	RF	0.786 ± 0.150	0.731 ± 0.258	0.634 ± 0.118
	NB	0.743 ± 0.141	0.728 ± 0.173	0.692 ± 0.121
	SVM	0.682 ± 0.259	0.686 ± 0.288	0.590 ± 0.191
	KNN	0.748 ± 0.149	0.760 ± 0.200	0.671 ± 0.121
	XGBoost	0.761 ± 0.163	0.752 ± 0.218	0.659 ± 0.125
	DNN	0.753 ± 0.179	0.722 ± 0.237	0.626 ± 0.132
	Mean	0.746 ± 0.035	0.730 ± 0.026	0.645 ± 0.036
MACCS	RF	0.751 ± 0.166	0.732 ± 0.257	0.613 ± 0.121
	NB	0.724 ± 0.142	0.720 ± 0.177	0.662 ± 0.117
	SVM	0.670 ± 0.253	0.681 ± 0.292	0.577 ± 0.190
	KNN	0.719 ± 0.147	0.750 ± 0.201	0.646 ± 0.119
	XGBoost	0.739 ± 0.168	0.741 ± 0.224	0.639 ± 0.124
	DNN	0.705 ± 0.181	0.697 ± 0.249	0.591 ± 0.121
	Mean	0.718 ± 0.028	0.720 ± 0.027	0.621 ± 0.033
Morgan	RF	0.774 ± 0.166	0.722 ± 0.282	0.612 ± 0.122
	NB	0.772 ± 0.143	0.745 ± 0.176	0.702 ± 0.124
	SVM	0.680 ± 0.268	0.685 ± 0.292	0.594 ± 0.192
	KNN	0.755 ± 0.154	0.755 ± 0.211	0.674 ± 0.124
	XGBoost	0.761 ± 0.164	0.749 ± 0.223	0.653 ± 0.128
	DNN	0.761 ± 0.176	0.715 ± 0.245	0.621 ± 0.132
	Mean	0.751 ± 0.035	0.729 ± 0.027	0.643 ± 0.041
PharmacoPFP	RF	0.757 ± 0.174	0.735 ± 0.258	0.620 ± 0.121
	NB	0.726 ± 0.144	0.722 ± 0.174	0.670 ± 0.123
	SVM	0.684 ± 0.240	0.689 ± 0.281	0.587 ± 0.184
	KNN	0.740 ± 0.147	0.761 ± 0.193	0.664 ± 0.120
	XGBoost	0.748 ± 0.175	0.745 ± 0.225	0.649 ± 0.129
	DNN	0.735 ± 0.183	0.709 ± 0.249	0.614 ± 0.130
	Mean	0.732 ± 0.026	0.727 ± 0.026	0.634 ± 0.032

^a RF: Random forest
^b NB: Naïve Bayesian
^c SVM: Support vector machine
^d KNN: K-Nearest Neighbor
^e XGBoost: Extreme gradient boosting
^f DNN: Deep neural networks
^g AUC: Area under the receiver operating characteristics curve
^h F1 scores: F1-measure
ⁱ BA: Balanced accuracy. “ ± ” values represent standard deviations

Back to article page

ISSN: 1758-2946

Contact us

Submission enquiries: journalsubmissions@springernature.com