Joint deep learning for batch effect removal and classification toward MALDI MS based metabolomics

BMC Bioinformatics

Table 3 MMDs of the SLE data before and after being calibrated by individual methods

Source	Target	Raw	ComBat	Ratio_G	fSVA	ResNet	NormAE	Ours	In Source
1	2	0.217 ± 0.010	0.125 ± 0.008	0.217 ± 0.009	0.285 ± 0.013	0.153 ± 0.010	0.202 ± 0.012	0.071 ± 0.008	0.053 ± 0.005
	3	0.696 ± 0.016	0.134 ± 0.009	0.292 ± 0.008	0.351 ± 0.014	0.144 ± 0.008	0.166 ± 0.008	0.072 ± 0.007
2	1	0.217 ± 0.010	0.125 ± 0.008	0.221 ± 0.010	0.214 ± 0.011	0.153 ± 0.010	0.202 ± 0.012	0.073 ± 0.008	0.062 ± 0.005
	3	0.623 ± 0.016	0.147 ± 0.009	0.245 ± 0.010	0.242 ± 0.010	0.176 ± 0.010	0.145 ± 0.008	0.069 ± 0.006
3	1	0.696 ± 0.016	0.134 ± 0.009	0.291 ± 0.008	0.289 ± 0.012	0.144 ± 0.008	0.166 ± 0.008	0.074 ± 0.008	0.064 ± 0.004
	2	0.623 ± 0.016	0.147 ± 0.009	0.240 ± 0.010	0.223 ± 0.008	0.176 ± 0.010	0.145 ± 0.008	0.070 ± 0.007

The first two columns present the data combinations of different batches participating in the comparison. The other columns are similar to the comparisons in Table 1

ISSN: 1471-2105