Table 2.

Performance of models with different ML algorithms and the concatenated descriptors

Descriptor	Method	Train AUC	Valid AUC	Test AUC	Accuracy	Precision	Recall	F1-score
F46/UniRep	GBM	1.000 ± 0.000	0.624 ± 0.163	0.834	0.759	0.751	0.706	0.728
	LGBM	0.928 ± 0.006	0.638 ± 0.166	0.830	0.746	0.719	0.729	0.724
	RF	1.000 ± 0.000	0.594 ± 0.174	0.823	0.747	0.742	0.683	0.711
	XGB	0.971 ± 0.002	0.624 ± 0.166	0.826	0.752	0.728	0.729	0.729
F46/TAPE	GBM	1.000 ± 0.000	0.669 ± 0.137	0.829	0.748	0.745	0.680	0.711
	LGBM	0.919 ± 0.006	0.676 ± 0.147	0.826	0.746	0.720	0.726	0.723
	RF	1.000 ± 0.000	0.654 ± 0.153	0.823	0.749	0.752	0.671	0.709
	XGB	0.997 ± 0.000	0.668 ± 0.151	0.831	0.748	0.727	0.718	0.722
F46/ESM-1b	GBM	1.000 ± 0.000	0.643 ± 0.154	0.830	0.755	0.746	0.704	0.724
	LGBM	0.882 ± 0.008	0.654 ± 0.157	0.821	0.744	0.709	0.746	0.727
	RF	1.000 ± 0.000	0.622 ± 0.169	0.826	0.750	0.748	0.680	0.713
	XGB	0.954 ± 0.004	0.648 ± 0.153	0.823	0.750	0.721	0.738	0.729
F46/ESM-1v	GBM	0.985 ± 0.002	0.643 ± 0.139	0.814	0.738	0.741	0.654	0.695
	LGBM	0.834 ± 0.010	0.657 ± 0.134	0.802	0.727	0.677	0.766	0.719
	RF	1.000 ± 0.000	0.611 ± 0.165	0.827	0.751	0.750	0.681	0.714
	XGB	0.951 ± 0.003	0.643 ± 0.151	0.822	0.743	0.709	0.742	0.725

The bold means the best performance, the AUC score in the test set