Table 3.

Comparisons between our models and other previous predictors of the MGMT methylation status in glioblastoma multiforme.

Study	Year	No. of features	Classifier	SN	SP	ACC
Le et al. 23	2020	9	XGBoost	0.88	0.887	0.887
Xi et al. 21	2018	63	Support vector machine	0.888	0.838	0.866
Levner et al. 45	2009	8	L1-regularized neural networks	0.854	0.9	0.877
Korfiatis et al. 40	2016	4	Support vector machine	0.803	0.813	N/Aa
Crisi et al. 42	2020	14	Multilayer perception	0.75	0.85	N/A
Kanas et al. 46	2017	N/A	K-Nearest Neighbor	0.736	0.852	0.663
Sasaki et al. 44	2019	5	LASSOb	0.67	0.66	0.67
L Han et al. 47	2018	N/A	CRNNc	0.67	0.67	0.67
Ahn et al. 43	2014	N/A	Mann–Whitney U-test	0.563	0.852	N/A
Our present study	2022	25	GA-RFb	0.894	0.966	0.925

SN, sensitivity; SP, specificity; ACC, accuracy.

^a“N/A” means that the information was not shown in the research.

^bLASSO, least absolute shrinkage and selection operator; GA-RF, genetic algorithm-random forest. Bold font indicates the results of this study.

^cBi-directional convolutional recurrent neural network architecture.