May I know what’s the purpose of performing qiime diversity procrustes-analysis? I have read through the help documentation, yet found confusing. Can anyone explain this plug-in with the provision of practical instance in microbiome analysis?
Procrustes is a really good visualization tool for comparing the patterns in two variants of your data.
My most frequent use is when I do data filtering. I tend to do feature-based analysis on a filtered subset of my data, where Im working with ASVs that are both abundant and prevelant at a certain threshhold. This might mean that I go from a table of 6000 ASVs to a table of 300. If I see global patterns with the larger table, it’s important to me to make sure those patterns still hold before I go forward with feature-based analyses or talk about conclusions in my paper. So, I tend to run a procrustes analysis to see if the global patterns still hold. Then, I’ll accompany it with a mantel test, because life is just better when you qualify things.
A few other concrete examples Ive used are times when I was comparing replicates with paired methods (again, a nice way to see if global relationships still hold spatially).
@jwdebelius, thank you for your kind reply. Does it mean this visualization was used to verify your filtration at a promising level by comparing the filtered subset feature table against initial feature table? If you mean so, may I ask question pertaining to technical code while running this test:-
i- reference-pcoa FILTER or INITIAL PCOA?
i-other-pcoa FILTER or INITIAL PCOA?
Then, you stated the output of this visualization abled to see whether the global patterns still hold. I apologize for my ignorance on the above-mentioned. Do you mean you compare your filtered data to the open-reference database? May I know more about this?
Apart from that, may I know in details on how to compare replicated with paired methods by using Procustres? Currently, I have 5 replicates of sample, is it comparing one by one?
I’m novice to microbiome world, may I know how to interpret the result from this analysis? As far as I know, there is a line connecting 2 dots on PCoA, what does it mean?
Yeah, I mean, it’s a pretty picture for my co-authors and reviewers who usually like pretty pictures. But, its also a good check for myself to make sure things look the way I expect. Ive never found the order of reference to filter to matter, as long as I know which is which going into the visualisation.
In this case, “global patterns” refers to the high level community level patterns in full dataset (As opposed to feature-based testing). Most of the analyses I dont likely won’t make sense in a larger landscape and Im focused on my sample set. There are some more global tools for reference, but that’t probably beyond the scope of this question and discussion. Its also probably not directly helpful to the problem, since PCoA space shifts with the introduction of new samples. So, it wouldn’t be a good approach for procrustes.
I would discourage using Procrustes in this case, it’s not going to test the hypothesis you want to test. I would, instead, suggest running a vanilla PCoA and coloring by source sample (vs replicate). Here, you’re checking if the replicates are more similar to each other than they are to replicates taken from other sources. You should see tight clustering around the points. Then, I’d run a permanova or adonis to check whether the replicates are more similar/cluster more tightly than the between sample distance.