I'm new to the forum and trying to figure out a few things regarding abundance testing. I have 2 groups (test and control n=8, n=7 respectively) and I see significant differences between the 2 groups in terms of alpha and beta diversity analysis as well (Unweighted p-value 0.003, weighted p-value 0.024 with PERMANOVA) with separation on the emperor plots. My problem is when I try to use ANCOM to look at the differential abundance, I don't get any significant results. I tried filtering at different parameters as well but that didn't work either. When I tried the differential abundance test at a specific taxonomic level, it gives me 2 families with W score of 18,19. Another differential abundance method I tried using was q2dsfdr which gave me multiple taxa that were different but since it only gives out the p-values I can't ascertain which group the taxa belongs to.
So, I have a few questions regarding the differential abundance testing:
Is it possible that although my alpha and beta diversity are significantly different among groups, I just don't have enough power to use ANCOM to actually pin point which specific groups are different ?
Can I consider the results given by ANCOM specific taxonomic level analysis to be valid?
Are the results from q2 dsfdr valid and if so how can they be processed for me to actually find out which group the different taxa belong to? **Also, I tried using q2dbbact as well but didn't get any results from that either.
I'm not sure what is actually going on with my data but I would really appreciate some help. Also, if there are any other differential abundance testing methods I can use.
Welcome back to the forum! Apologies for the delay in response on this.
Yes! This can absolutely be the case. You may just not be seeing significant results at the ASV level, which is not uncommon. You may see more significant results by zooming out, say, to the genus level.
Yes!
The pattern and differential abundance might only emerge at the genus level.
I would ask @serenejiang about the validity since they are the plugin developer who created q2-dsfdr; it looks like this plugin hasn't been maintained over the last few years. That being said, you may just not be getting significant results because there actually aren't any for your dataset at this level.
These are a couple of other community plugins that can be used for differential abundance testing that are currently maintained:
Thank you so much for your response. This is really helpful. I tried to move my data to R and use ANCOM-BC for analysis which does show some groups that were differentially abundant but I'm not sure if I was supposed to filter the sequences before using that or not. Another analysis tool I used is LEFse which also showed taxa that were differentially abundant. I am just confused as to which method is more reliable as I do see some common taxa among these different analysis tools but the overall results are different. I would really appreciate if you could please help elaborate on this a little.
Thank you.
There will be some variability between tools - the variability will also differ between tools depending on the dataset, and whether or not you are filtering prior to running DA. That being said, @jwdebelius has a paper in pre-print that has shown the following:
Limma voom, Wilcoxon, LEfSe, and edgeR tended to find the largest number of significant ASVs compared with other methods.
I would highly recommend reading her paper (linked here), as it will help clarify your questions/concerns in much greater detail than I can.
It's not actually my paper! I nominate papers for the most impactful paper/preprint of the year , and tI think this one will push the field forward. But yes, highly recommend it
forum! Apologies for the delay in response on this.
, and tI think this one will push the field forward. But yes, highly recommend it