Table 1. Evaluation results of four criteria on benchmark dataset MTBLS79 (selected measure under each criterion was shown in bracket).

	Criterion (a)	Criterion (b)	Criterion (c)	Criterion (d)
	(PMAD)	(distribution of P-value)	(consistency)	(AUC)
Auto scaling	0.8360	Good	14.6500	0.8344
Contrast	0.7797	Fair	9.7500	0.6250
Cubic splines	0.1393	Excellent	13.7500	0.8322
Cyclic loess	0.3188	Good	15.6500	0.8356
EigenMS	0.1799	Good	16.4000	0.8010
Level scaling	0.2890	Good	15.1000	0.8345
Linear baseline	0.6035	Fair	6.3000	0.7072
Log-transform	0.1349	Good	14.7500	0.8168
Mean	0.3100	Good	14.7500	0.8213
Median	0.3100	Good	14.5500	0.8177
MSTUS	0.0064	Good	14.3500	0.8405
Pareto scaling	0.5320	Good	14.9500	0.8344
Power scaling	0.1660	Good	14.9500	0.8314
PQN	0.3260	Good	13.7000	0.8309
Quantile	0.2989	Excellent	13.8000	0.8119
Range scaling	0.1573	Good	15.3500	0.8344
Total sum	2.4336	Fair	14.7000	0.7538
Vast scaling	2.7200	Good	15.0000	0.8344
VSN	0.5626	Excellent	13.7500	0.8373

The way calculating those measures was described in ‘Materials and Methods’ section and ‘Supplementary Methods’ section. Besides of quantitative measures, qualitative ones such as distribution of P-value were also evaluated and three performance levels were provided (Excellent, Good and Fair). Qualitative measures were evaluated by visual inspection and examples illustrating how those three performance levels were assigned were shown in Supplementary Figure S1.