Skip to main content

Table 3 Untargeted features selected as predictors of case-control status

From: Untargeted lipidomic features associated with colorectal cancer in a prospective cohort

Feature ID

Observed m/z a, b

Ret. time (sec)

Fold change

p-value

Feature type c

235

391.2832

596.4

0.702

0.000261

R

4250

453.3592

605.5

0.753

0.001082

R

4294

467.3744

605.6

0.741

0.000569

R

5080

519.1965

595.9

1.308

0.000432

C

3207

531.1558

563.9

0.661

0.000468

C

6054

551.1781

563.9

0.404

0.001750

C

839

577.2698

620.6

1.370

0.000359

C

5749

882.6393

718.2

0.880

0.000168

I

14,963

907.4806

617.6

1.373

0.000133

R

  1. m/z mass-to-charge ratio, p-value from the regression model (Eq. 2), C potentially causal feature, R potentially reactive feature, I indeterminate
  2. aObserved m/z values correspond to singly-charged negative ions
  3. bFeature selected by bootstrap LASSO and by being in the top 1% of features ranked by both the p-values from the case-control regression (Eq. 2) and the random forest variable importance measure
  4. cBased on regression of case-control difference on time to diagnosis (ttd, Fig. 2)