Interpreting aldex2 Results

Community Plugin Support
1

I'm able to run everything fine and get plots that indicate two features are significantly different (appear red in the volcano plots). However, when I do the next part of the tutorial to extract these features, it appears they are not significant. If I run:

qiime aldex2 extract-differences \
    --i-table gut-test/differentials.qza \
    --o-differentials gut-test/sig_gut \
    --p-sig-threshold 0.1 \
    --p-effect-threshold 0 \
    --p-difference-threshold 0

I get the error: " You have selected a significance threshold that is lower than minimum Q score (-p--sig-threshold). Select a higher threshold."

I have to bump up the sig threshold to 0.45 to get the two features. Does this mean they are actually not significant?
differentials.qza (244.7 KB) sig.qza (203.8 KB) viz.qzv (445.8 KB)

edit: tagging @dgiguer

1 Like
2

It looks like it depends on if you use the t-test results (we.eBH) or the Wilcoxon/mann-whitney test (wi.eBH):

Feature.ID rab.all rab.win.control rab.win.vancomycin.1.day diff.btw diff.win effect overlap we.ep we.eBH wi.ep wi.eBH
TACGGAGGATGCGAGCGTTATCCGGATTTATTGGGTTTAAAGGGTGCGTAGGCGGTTCGTTAAGTCAGCGGTAAAATTGCGGGGCTCAACCCCGTCGAGCCGTTGAAACTGGCAGACTTGAGTTGGCGAGAAGTACGCGGAATGCGCGGT 1.324030 -0.7753869 10.921798 11.41690 3.246717 3.318627 0.0002193 2.48e-05 0.0047024 0.000999 0.1667048
TACGGAGGATGCGAGCGTTATCCGGATTTATTGGGTTTAAAGGGTGCGTAGGCGGACTGTCAAGTCAGCGGTAAAATTGAGAGGCTCAACCTCTTCCCGCCGTTGAAACTGGTGGTCTTGAGTGGATGAGAAGTACGCGGAATGCGTGGT 0.328773 -1.3907614 8.772128 10.32092 2.722074 3.909851 0.0002193 1.55e-05 0.0032193 0.000999 0.1667048
1 Like
3

As @jbisanz mentioned, it depends on the t-test that you use. Determining whether there is a true biological differences requires some interpretation using the context of your dataset. You will have to think about the nuances of your dataset to decide whether there is truly a biological difference or not.

Cheers,

Dan

4

Hi Dan,

So it appears the plots highlight with the raw p-values then while the “extract differences” function is going off the “we.eBH” (I assume that is benjamini hochberg corrected p-vals?). Because neither of my BH corrected pvals (we.eBH) or the Wilcoxon/mann-whitney test (wi.eBH) are below 0.1…

I assume we would never want to use the raw pvalues in any case

1 Like
5

Hi @vrbana,

You can find more information about what each of the outputs is the ALDEx2 vignette. You are correct in we.eBH is the BH-corrected values.

Dan

6

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.