Table 3.

Binary classification results of baseline models for nonharmonized and harmonized data. For Parkinson’s disease and schizophrenia, results comparing the accuracy of different models for nonharmonized and harmonized methylation data are shown.

Model	Parkinson’s disease		Schizophrenia
	GSE72774		GSE152027		GSE116379
	Nonharmonized	Harmonized	Nonharmonized	Harmonized	Nonharmonized	Harmonized
Logistic regression	0.71	0.93	0.63	0.66	0.56	0.66
Support vector machine	0.67	0.92	0.62	0.66	0.58	0.65
XGBoost	0.72	0.95	0.67	0.71	0.56	0.66
CatBoost	0.71	0.94	0.68	0.72	0.59	0.71
LightGBM	0.76	0.97	0.68	0.71	0.58	0.67
TabNet	0.69	0.93	0.63	0.66	0.58	0.65
NODE	0.71	0.92	0.62	0.66	0.56	0.65