Good morning to all!
I have a doubt about longitudinal analyses performed on my samples. I will explain:
I have 120 fecal samples (collected from 60 animal divided in three groups of 20) fed with different diet and collected in two different time T0 and T1.
The taxonomy suggest me (and this data is consistent) that between two paired samples there is a difference in the microbiota composition, but when I performed the longitudinal analyses the result showed me not significant difference. I want to know if the diet impact on the microbiota.
The script is the follow:
qiime longitudinal pairwise-distances
–i-distance-matrix weighted_unifrac_distance_matrix.qza
–m-metadata-file metadata_vitelli.tsv
–p-group-column Diet
–p-state-column Time
–p-state-1 0
–p-state-2 1
–p-individual-id-column Subject
–p-replicate-handling random
–o-visualization wei_dist.qzv
Someone can help me to understand if it is correct the script?
thanks in advances
Hi @rparadiso ,
It looks like you are setting up your test correctly and understanding it correctly — so we cannot argue with the statistical results (!), but just to comment:
Qualitative taxonomic results and weighted unifrac based on ASVs are different views of the same data and will not necessarily agree. E.g., weighted UniFrac is going to weight by abundance and phylogenetic distance, so would be less sensitive to, e.g., differences in taxa that are phylogenetically closely related. You could perform the same test with different distance metrics, or even Bray-Curtis distance based on taxonomic composition, and see if there is a distance. These metrics all give different "views" of the data and comparing these results can be instructive, even (or especially) if one metric gives a significant result and the other does not.
Good luck!
1 Like
Hi @Nicholas_Bokulich
Thank you so much for your consideration and your help!
Regards
1 Like
Thank you Nicholas for your kind reply.
However I still have a doubt.
In my longitudinal analysis I can see that the three groups all change in a similar way (indeed I have no significant p-values when comparing the three T1 times), but I can see that for all the three groups, I have a significant change between T0 and T1.
I would like to know if there is a way to measure this differences and to evaluate the significance of my data, e.g. I want to know if I can say that there is a significant difference, within each group, between T0 and T1, even though there are no significant differences among groups at T1.
I look forward to receiving your reply
thanks a lot
Rubina
Hi @rparadiso ,
Only pairwise-differences tests that the difference between states is > 0. pairwise-distances does not because distances between samples are almost always > 0 and cannot be < 0, so will always be significant (unless if samples are exactly the same, which is extraordinarily unlikely for microbiome datasets). Thus such a test would be quite meaningless and so is not tested here.
Good luck!
