Table 2.

Wilcoxon-Mann-Whitney two-sample rank sum test applied to screened, quantile-normalized median summarized data from the Nickel Exposure Study. The samples contain 18 subjects with high (≥ 9.98 μg/L) and 18 subjects with low (≤4.44 μg/L) level of urinary Nickel. The meaning of columns are as follows: GID – glycan identification number, Z – z-statistic, p – p-value of the test, FDR – false discovery rate, AUC – area under the ROC curve, AUC_C – cumulative AUC value obtained by combining the above glycans through multivariate logistic regression, ICC – corresponding Intra-class Correlation Coefficient computed for raw data. The sign of the z-statistic indicates downregulation (negative sign) or upregulation (positive sign) of normalized signals. The low values of FDR, at least for top three glycans, imply a good confidence in the results. The glycans in the table are sorted by ascending order of the p-value. The sixth column suggests that combining several glycans can considerably increase the AUC value. For example, combining three top glycans: GID = 191, 264, 133, gives AUC_C = 0.966. Figure 13 shows a solid statistical significance for this AUC value.

GID	Z	p	FDR	AUC	AUCc	ICC
191	3.797	0.00015	0.0181	0.8704	0.8704	93.3
264	−3.497	0.00047	0.0353	0.8457	0.9167	89.1
133	−3.149	0.00164	0.0584	0.8148	0.9660	88.7
136	−3.058	0.00223	0.0602	0.8117	0.9722	82.2
135	−2.959	0.00309	0.0756	0.7932	0.9784	93.3
384	−2.927	0.00342	0.1490	0.7901	0.9815	94.5
134	−2.897	0.00377	0.0668	0.7901	1.0000	81.7
93	−2.864	0.00418	0.0629	0.7840	1.0000	91.5
379	−2.864	0.00419	0.1186	0.7840	1.0000	94.5
137	−2.792	0.00524	0.0807	0.7932	1.0000	88.9