ancom results has significant but very low/0 W value

I’m very beginner of metagenome analysis…
I used qiime2 Ancom and found that some taxa with very low or 0 W value was significant results.

Could anybodyancom_female-N4-L2.qzv (631.8 KB) see my data and tell me if I can ignore that odd significant taxa ?
or I have to reanalysis with ancom R ?

Thanks

1 Like

Hi @eunkyo85,
It looks as though you using ANCOM at the phylum level. From my own experience and maybe someone can confirm, ANCOM is much more reliable when you are dealing with more features as the maximum W score you can get at the phylum is pretty low, in your case 16-1=15. I also don’t see any indication that even that 1 taxa is considered significant, but with a W=4, it’s hard to be sure really.
I wouldn’t trust the 1 taxa story without some additional testing. Try your data at the Genus or ASV level which taps into the strength of ANCOM and you might see some other real differences.

2 Likes

Thanks for advice. I used Ancom at another levels also.
at class level, only D_0__Archaea;D_1__Euryarchaeota;D_2__Methanobacteria with a W=7 considered significant. and at the other levels (family, order, genus, species), there wasn’t any significant data.

Hi @eunkyo85,
Then it would stand reason to believe perhaps there are no differences between your groups. At least according to ANCOM, which has its own considerations. For example, reducing noise from your table by removing rare features or those occurring in only a few samples can enhance the ANCOM tests. You can also use other tests to confirm these findings with some other tools like gneiss, LEfSe, corncob etc.
Also, if the Metanobacteria has an any implications within your project you could look into that group on its own further since it was identified as potentially important, though again a W=7 doesn’t really strike me as something I would trust right away.

1 Like

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