Pairwise comparison between samples


(Ben) #1

Hello,

First, thank you for making this forum available to users. I have learned a great deal from it. My question is:
I have 7 samples from the same environment, labeled 1-7 in the metadata file. I don’t have groups or treatment or anything like that. What Qiime 2 command can I use to do a pairwise comparison of these samples to see if there is a significant difference in taxa abundance between them? Can I use the longitudinal pairwise-differences command for this?

Thanks for your help.

Ben


(Mehrbod Estaki) #2

Hi @ascorsel,
Without a group/treatment assignment what you are asking is statistically not possible/meaningful. Consider you compare the height of a male (170cm) and a female (173 cm) student from a classroom. You can say empirically that their heights are different from each other but you cannot make any generalization to the rest of the population based on this single observation. But if you were to sample 10 females and 10 males from the class, then you can compare the means of these groups and make some generalization about the rest of student population in that class. The same logic applies here where you are not assigning any groups/treatments to your samples. You can look at the microbial community of these 7 samples (i.e. alpha diversity, PCoA plots, composition barplots etc.) and make empirical observations but you are’t able to test any hypothesis which is required for stats. That means you can’t say this sample is statistically different than another.


(Leo CA) #3

Hi Mehrbod,
I agree with you. There should be at least control vs treatment samples. @ascorsel, is there at least any divergence between those samples out of the environment, such as time of sampling? Otherwise they look like actually a lot of replicates from the same condition.

@mehrbod_estaki Just curious here: it is ok to run pairwise diversity analyses for control vs treated samples with no replicates? More specifically, do people often do that and it’s acceptable for microbiome studies or maybe it’s just a matter of having or not enough individuals for replicating? =)
Thanks!


(Mehrbod Estaki) #4

Hi @lca123,
If you are referring to biological replicates then the answer is the same as above, you can’t actually compare variance across groups if there is no variance (i.e. one sample in a group). If you are referring to technical replicates however, that’s a whole other story. There is already a discussion regarding technical replicates happening here which I think is where any further discussion should jump to :+1:


(Leo CA) #5

Many thanks! Lucky there is a fresh thread on that.