Significant Features in ROC Analysis

[szhao1]

Hi,

I am trying to use the Biomarker Analysis module for multivariate ROC analysis. I used the demo dataset 1 on the website and then tried several times using the exact same parameters. However, I found that the red/blue box beside each metabolite in the Sig. Features are different each time I tried. Please see images below as I tried two times.

I am wondering what is the meaning of the red/blue box and does it represent the concentration? For example, in the demo dataset 1, I can see that glycerol is significantly higher in 0 group than 1 group. In the images above, sometimes it showed as red in 0 and sometimes it showed as blue in 0 group. Thank you!

[jeff.xia]

Please read our response to this question, and update your post.

[szhao1]

Hi Jeff,

Thanks for the instruction. Below please see more information about the trial.

1. Tool and Module: MetaboAnalyst 5.0, Biomarker Analysis

2. Data: Example Dataset 1 (Metabolite concentrations of 90 human plasma samples measured by 1H NMR. Phenotype labels: 0 - Controls; 1 - Patients.)

3. Step: The parameters used in each step are:
   a) Missing values: just click proceed

b) data normalization: as below

c) functions: as below

d)classification method: as below

The same procedures were repeated several times. However, even using the same parameters, different results can be generated, as below

For example, for glycerol, sometimes it showed as red box in 0 group, while sometimes it showed blue box in 0 group. In the data spreadsheet, 0 group has higher amount than 1 group for glycerol.

Thank you!

Best,

[jeff.xia]

Thanks for the details! I can reproduce the issue. It should be fixed now.

For clarification, the color of the mini heatmap for each biomarker is based on its median values in each group to show its relative changes.