Thank for your attention.
There are many pathways that we would expect to have changed in gut microbiome before and after the diet intervention. However, through pair Wilcox test, we detect no biological meaningful pathway based on KEGG and eggNOG data.
Might we should try some less strict method?
What is your opinion?
You can't get less strict than a Wilcox test without multiple test correction. Note that the Wilcox test is not really appropriate for testing for significance in microbiome data, as the compositionality of microbiome data breaks some of the test's assumptions and it will be prone to a high false-positive rate.
Maybe try some of the plugins available in QIIME 2, like an ANCOM test (via the q2-composition plugin). Check out the tutorials at qiime2.org for some examples.
Good luck!
2 Likes
I’m going to add a different opinion that Nick’s. If you’re not seeing a difference statistically, it may be that there’s not one there. Or that you’re underpowered to see it. Think about the dimensions of your data (number of samples vs number of features) and why biologically you think there should be a difference.
Do you have evidence for changes in pathways based on diversity analyses?
If you’ve done your due dilligance and your sample size is large enough and you don’t see a difference, chances are the answer is that there is no difference. If your sample size is too small, your answer ist hat you can’t answer the question with your current sample size. In my experience, it’s frequently the later.
3 Likes
Shall composition effect in metagenomics be weakened by the large number of species? About differential test, what is your opinion on Deseq2?
Hi @Wang_cs001632,
The compositionality of your data is not affected by the number of features present. The number of features affects your multiple hypothesis correction and your power (unfortunately) and so filtering is a way to cheat.
DeSeq2 isn’t appropriate for compositional data in the microbiome.
Best,
Justine
1 Like