Table 2.

Comparison of solubility prediction performance of PLM_Sol with existing models on independent test set.

Models	T	AUC	Accuracy	F1_score	MCC	Precision	Sensitivity	Specificity
Protein_sol	0.5	0.5985	0.5649	0.6350	0.1405	0.5470	0.7569	0.3729
SKADE	0.5	0.6882	0.6443	0.5304	0.3305	0.7811	0.4015	0.8874
SWI	0.5	0.5597	0.5423	0.6334	0.0971	0.5284	0.7905	0.2938
Soluprot	0.5	0.7126	0.6308	0.6639	0.2667	0.6095	0.7290	0.5325
EPSOL	—	0.6664	0.6663	0.5915	0.3578	0.7630	0.4830	0.8499
NetSolP	0.5	0.6183	0.5502	0.6557	0.1268	0.5312	0.8564	0.2437
DeepSoluE	0.4	0.6660	0.5990	0.6407	0.2036	0.5804	0.7150	0.4830
PLM_Sol	0.5	0.8342	0.7299	0.7542	0.4690	0.6919	0.8289	0.6308

T represents the threshold value of the models. The bolded string indicates the highest score within the respective column.