Beta_diversity.py and Beta_significance.py

[Cláudia Vicente]

Hello Qiime Forum,

My post is related with output from Beta_diversity.py and Beta_significance.py. I am studying the effects of a parasitic nematode in the gut-microbiome of an insect. I have prepared six 16S rRNA libraries for respectively: With nematode foregut, midgut and hindgut; and Free-nematode foregut, midgut and hingut. I have analyzed beta_diversity using distance and phylogenetic matrices. Regarding the phylogenetic matrices, ANOSIM test with weighted (test statistics: -0.22; p-value: 0.899) and unweighted UniFrac (test statistics:0.88; p-value: 0.107) resulted in no statistical differences between with nematode/free-nematode categories and body parts. However if I run beta_significance.py with weighted_normalized_unifrac/weighted_unifrac/unweighted_unifrac and test each pairwise samples, I can see statistical differences between some samples. Which results can I trust?

Moreover, if I analyze by t-test with unweighted unifrac, nonparametric p-value Bonferroni corrected is statistically significant (within vs between categories - with nematode and free-nematode) and within category vs with nematode and free-nematode. 

I would very grateful if someone can help me answering or referring some literature that can help me with this subject. 

Thank you. 

[zech xu]

Hello,

How did you do the ANOSIM tests? you divided your 6 samples into 2 groups, 3 with-nematode and the other 3 free-nematode? If so, all the results are valid and the two sets of tests you did are testing different things. ANOSIM tests if there is difference between 2 groups. But you have heterogeneity in each group (different gut sites), which may lower your statistic power. beta_significance tests if sample A is different from sample B by permutation tests.

Best,
Zech

[Cláudia Vicente]

Hello Zech,

Thank you very much for your answer. Indeed, I divided my 6 samples in groups of 2 and the ANOSIM. I will use then the beta_significance test output.