Metabolomics variable selection and classification in the presence of observations below the detection limit using an extension of ERp

Table 1 XERp Results for TBM vs Healthy Controls

Variable	ER	C	Direction	Observed p-value	Percentage zeros
Variable	ER	C	Direction	Observed p-value	Control	Experimental	Overall
2-hydroxybutyric acid	0.08	0.3	Up	0	68%	0%	44%
3-hydroxyisovaleric acid	0.09	20.05	Up	0	0%	0%	0%
4-hydroxyphenylpyruvic acid	0.11	0	Up	0	100%	18%	71%
methylcitric acid	0.1	0.45	Up	0	55%	6%	38%
quinolinic acid	0.05	2.38	Up	0	45%	0%	29%
2-hydroxyvaleric acid	0.12	0	Up	0	77%	6%	52%
non-annotated-1	0.15	0	Up	0	100%	24%	73%
uracil	0.13	0.67	Up	0	48%	0%	31%
1,4-dihydroxycyclohexane	0.14	0.23	Up	0	55%	6%	38%
non-annotated-3	0.19	0	Up	0	100%	29%	75%
2-ketoglutaric acid	0.16	0.56	Up	0	55%	6%	38%
phenylacetylglutamine	0.17	1.02	Up	0	90%	24%	67%
hexanoic acid	0.16	0.34	Down	0	35%	35%	35%
pyruvic acid	0.17	0.25	Up	0	29%	0%	19%
isocitric acid	0.17	1.6	Up	0	32%	0%	21%
glycolic acid	0.17	12.98	Up	0	0%	0%	0%
pyroglutamic acid	0.17	4.69	Up	0	52%	18%	40%
vanillylmandelic acid	0.18	7.5	Up	0	13%	0%	8%

The table lists the discriminatory variables based on their significant \( {p}_{obs} \)-values when compared to B-H adjusted significance level (\( \alpha =0.05 \)). The first column lists the names of these variables (labelled “Variable”), the second the minimized error rate (labelled “ER”) and the third the associated threshold (labelled “C”). The directions of the minimized error rates are given in the fourth column labelled “Direction”. Column five contains the \( {p}_{obs} \)-values. The last three columns also report the percentages of zeros in each group and in the combined data set

ISSN: 1471-2105