Hello everyone,
I have a 16S rRNA dataset of the gut microbiome of 11 humans, who were given two different dietary interventions. For beta diversity analysis I did ANOSIM of the Bray Curtis Dissimilarity, first with Qiime and later on with R. Now I get different results although I use the same dataset.
Qiime: test statistic: 0.2123 p: 0.001
R: test statistic: 0.0083 p: 0.298
I already checked the Bray Curtis Calculations, they are equal in Qiime and in R. Does anyone have a clue what might be the problem, or can anyone help me please?
I don't know how to trust the results now.
Thanks in advance!
So you double checked that's the output of the Bray-Curtis matrix is the same between Qiime and R (same values)? Did you ensure that's no samples labels were swap or are different between R and Qiime, the factor you use is not different?
What you are testing is if you have difference in the gut microbiome between humans having dietary A or B? So you have 11 humans microbiome, and 5 have dietary A and 6 dietary B or 11 dietary A and 11 dietary B?
Did you check if you run the same number of permutations, although I would not expect that to give completely different results IF Qiime and R do the same thing...
Did you run ANOSIM in both software not ADONIS in one and ANOSIM in the other?
Are you expecting significant difference with the ANOSIM? If yes, it could mean that Qiime work on a good set of data and not R. The only thing I can see is that if there is issues with your factor or samples labels. Hard to tell without seeing the code, but did you attach your factor to your OTU table in R? This can lead to mistake in matching samples and R would test based on wrong factor.
Hope that helps a bit
Hello Aimeric,
thank you for your fast answer!!
I checked the values and they were exactly the same and also number of permutations were the same. But after your comment I checked the R code again and found that the mapping file was melted which lead to a mismatching of the samples with the distance matrix. Running the pipeline again on my samples without melting the map, I get the same results as with Qiime. So it was not a calculation problem but a matching problem when running the ANOSIM command.
That solved my problem perfect.
Thanks!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package