Perform Diversity analysis

Dear all,

I performed diversity analysis using the following command lines

qiime diversity alpha
–i-table table.qza
–p-metric observed_otus
–o-alpha-diversity observed_otus_vector.qza

qiime diversity beta
–i-table table.qza
–p-metric braycurtis
–o-distance-matrix unweighted_unifrac_distance_matrix.qza
How can i convert the qza files to qzv because i want to visualize the results ?

The aforementioned command lines were used for non-phylogenetic diversity metrics.
What are the main differences between results obtained from phylogenetic diversity analysis and non phylogenetic diversity analysis What is the best method to perform diversity analysis phylogenetic or non phylogenetic approach ?

Thank you so much

Hi @M_F,

This is a complex question and one that, i think, is still a big point of discussion in the field. Ive had multiple conversations with people who swear by phylogenetic or non-phylogenetic metrics for their data. It depends on what you’re looking for in your data. It’s also worth keeping in mind that techniques can be complementary.

Imagine I have 3 communities:

  • Community 1: :mouse2:, :rat:
  • Community 2: :mouse2:, :rabbit:
  • Community 3: :computer_mouse:, :rat:

The non-phylogenetic metrics weights the similarity between any two features the same way, so the distance between 1 and 2 and 1 and 3 will be the same.

On the other hand, phylogenetic metrics account for the similarity between the features. Typically, this is thee shared evolutionary history - the phylogeny. (I’ll let you decide if you. think a :computer_mouse: is more similar to a :mouse2: than a :rabbit2: is to a :rat:… I think it depends on your perspective today.)

Best,
Justine

4 Likes

Hello @jwdebelius

Thanks for the response. For alpha and beta diversity analysis i obtained a qza files i would like to visualize those files and to convert them to qzv format How can i convert them ?
Thanks

Hi @M_F,

Have you looked at our fabulous tutorials?

Best,
Justine

Hello @jwdebelius,
I generated qzv files for alpha and beta diversity analysis.

qiime diversity alpha-group-significance
–i-alpha-diversity obersved_otus_vector.qza
–m-metadata-file sample-metadata.tsv
–o-visualization observed_otus_vector.qzv

For alpha diversity analysis using the aforementioned command line statistical analysis were performed automatically using Kruskal Wallis test/the last test don’t show significant differences between for example control group and Affected patients while in the boxplot i see significant differences i’m asking if it’s possible to replace the statistical test used (kruskal wallis) by another ?

  1. concerning beta diversity analysis i used the following command line
    qiime diversity beta-group-significance
    –i-distance-matrix unweighted_unifrac_distance_matrix.qza
    –m-metadata-file sample-metadata.tsv
    –m-metadata-column Disease
    –o-visualization core-metrics-results/unweighted-unifrac-subject-group-significance.qzv
    –p-pairwise

The plots generated using the aforementioned command line display for for example control group (n=1081) while the number of the control group is n=40 why the results show duplication of samples ?

  1. How can i generate PCOA results from beta diversity analysis in order to explore them using emperor tool ? i didn’t see the corresponding command line.

Thanks for your help

Hi @M_F,

You can't change the alpha group significance test because a normal distribution isn't always best for alpha diversity. I'm guessing you have relatively small sample sizes? I guess one of the questions I should have asked sooner is if you rarified your data?

Your beta diveristy questions have been addressed before in multiple threads, those are compiled into the Frequently Asked Question Post.

I think you'll probably find your answers there, but if not, please come back and ask more qustions!

Best,
Justine

Hi @jwdebelius
Thanks for the message.

  1. Regarding the alpha diversity analysis I didn’t perform alpha rarefaction plotting because as required by the following command line i should generate rooted-tree.qza file from the merged sequences however when i performed phylogenetic analysis of merged sequences of 100 samples i received an error message indicating that the job requires enough memory to achieve this task noting that the job was stopped because i don’t have enough memory. So i was obliged to conduct phylogenetic analysis running each sample separately (one by one) to create rooted-tree.qza file for each sample. The question is how to merge the rooted-tree.qza files created for each sample to perform alpha rarefaction plotting ?
    qiime diversity alpha-rarefaction
    –i-table table.qza
    –i-phylogeny rooted-tree.qza
    –p-max-depth 4000
    –m-metadata-file sample-metadata.tsv
    –o-visualization alpha-rarefaction.qzv

  2. Concerning the beta diversity analysis

i didn’t find a response to my questions in the FAQs :
The plots generated using the following command line show duplication of samples how can i resolve this problem (eg the number of the control group is n=40 i found n = 1081) ?
qiime diversity beta-group-significance
–i-distance-matrix unweighted_unifrac_distance_matrix.qza
–m-metadata-file sample-metadata.tsv
–m-metadata-column Disease
–o-visualization core-metrics-results/unweighted-unifrac-subject-group-significance.qzv
–p-pairwise
How can i generate PCOA results from beta diversity analysis in order to explore them using emperor tool ? i didn’t see the corresponding command line.

Thanks

Hi @M_F,

Can you please back up and describe your pipeline? You should not have a per-sample tree, so I'd love to know how you got there?

The number at the bottom of the boxplot is the number of distance measurements, not the number of samples.

Best,
Justine

1 Like

@jwdebelius,

Thanks for the reply.

  1. Regarding the beta diversity analysis :
    I have a question about the pairwise permanova results table what does those columns (permutations, pseudo-F, q value) refer to ?
    In boxplots we have y axis representing distance and in x axis we have for example group 1 (n = 1128) Group 2 (n=480) what does those n values represent if y axis represents distance ?

  2. Concerning the alpha diversity analysis :
    In the table representing Kruskal Wallis results : what did the column H refer to ?
    Thanks

Hi @M_F,

The pseudo-F is the test statistic for the original data. It's not used directly to calculate the p-value.
Permanova is a permutation test, so permutations tells you how many times the data has been shuffled to give you your p-value.
The q-value is the p-value adjusted for multiple hypothesis correction.

This was already discussed:

The H-value is the Kruskal wallis test statistics.

Best,
Justine

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.