Jackknifed Beta Diversity

[Niccolo`]

Hi everybody,

I have an issue regarding the Jackknifed Beta Diversity analysis. I am trying to give statistical support to my UPGMA tree and PCoA plots and I ran the following command:

jackknifed_beta_diversity.py -i otu_table_silvablast_no_singletons_no_chimeras_no_chloro.biom -o /jackknifed_beta_diversity_silva3 -e 121033 -m MK3-12mappingfile2.txt -t rep_set.tre

(121033 is the 75% of the sequences of the least abundant sample included in my analysis)

after running the command "make_bootstrapped_tree.py -m master_tree.tre -s jackknife_support.txt -o jackknife_named_nodes.pdf" I can see that the nodes of my tree are well supported 
(bootstrap > 0.9 for 6 out 7 nodes). (FILE 1)

Then I tried to compare the rarefied PCoA plots using emperor. I executed this command:

make_emperor.py -i /pcoa -m MK3-12mappingfile2.txt -o /jackknifed_pcoa -e sdev

Getting the file FILE 2: in this file I can see that the cloud of each sample is pretty narrow and there is never overlap in the clouds among samples.

Now I am trying to interpret such results. Since in the bootstrapped tree I can see that certain samples cluster together with good support (for example MK4 and MK6 or MK11 and MK12), 
I was expecting also the clouds of these couples of samples to overlap in the jackknifed 3D PCoA plot. But it is not the case.

I would like to ask if I made everything right in the commands I used and if I understood correctly the meaning of these analyses, or if I making any mistake in the 
interpretation of the results.

Is it right to think that if the samples cluster together with good support in the UPGMA tree, they should also be "close" in the PCoA plot and with overlapping clouds?

Thank you very much for any possible explanation and help.

Best regards

Niccolo'

[Niccolo`]

I forgot to mention that whenever I run make_emperor.py I always get the following message:

/usr/local/lib/python2.7/site-packages/emperor/format.py:101: FutureWarning: comparison to `None` will result in an elementwise object comparison in the future.
  if coords_low != None and coords_high != None:

the command runs anyway and I got the output, but I don't know if this is relevant to understand if there is something wrong.

Best regards

Niccolo'

[Sophie]

Hi Niccolo,

What version of numpy do you have?  You may also try updating your emperor version.  I don't think that warning will impact your results.

As for the jackknifed_beta_diversity.py analysis - I think it looks ok.  The spacings between the samples in the tree vs. the pcoa plot are similar.  Note that the 'clouds' around the PCoA points are a measure of uncertainty in that point's position, which is different from the sample placement on the tree.  Two samples next to each other on the UPGMA tree doesn't necessarily mean they are so similar that they will overlap on the pcoa plot - but mostly that those two samples are more similar to each other compared to the rest of the samples.

Hope that helps,

Sophie

[Niccolo`]

Hi Sofie,

I donwloaded the last versions of mumpy and emperor so I think that is not the problem. Anyway I also have the feeling that the warning has an effect on my results.

Thanks a lot for the explanation about the jackknifed PCoA plots and the UPGMA, now it looks more clear and understandable to me. I think I had misunderstood the meaning of the clouds in the PCoA.

I only have one more question:

Using the make_bootstrapped_tree.py command I generated the bootstrapped tree with the colours corresponding to different support values for the nodes.

Are the support values given bootstrap values or jackknife values? I guess they are jackknife values since jackknifed_beta_diversity.py perform jackknifing, but using the make_bootstrapped_tree.py made me rise the doubt.

Thanks a lot

Best regards

Niccolo´

[Yoshiki Vazquez Baeza]

Hello,

Just to clarify, that warning doesn't affect the results you get and is
going to go away once we release the next version of Emperor (0.9.6),
see this pull request: https://github.com/biocore/emperor/pull/331

Thanks!

Yoshiki.

> >>> otu_table_silvablast_no_singletons_no_chimeras_no_chloro.biom -o */*jackknifed_beta_diversity_silva3

> >>> -e 121033 -m MK3-12mappingfile2.txt -t rep_set.tre
> >>>
> >>>
> >>> (121033 is the 75% of the sequences of the least abundant sample included in my analysis)
> >>>
> >>> after running the command "make_bootstrapped_tree.py -m master_tree.tre -s jackknife_support.txt -o jackknife_named_nodes.pdf" I can see that the nodes of my tree are well supported
> >>> (bootstrap > 0.9 for 6 out 7 nodes). (FILE 1)
> >>>
> >>> Then I tried to compare the rarefied PCoA plots using emperor. I executed this command:
> >>>
> >>> make_emperor.py -i /pcoa -m MK3-12mappingfile2.txt -o /jackknifed_pcoa -e sdev
> >>>
> >>> Getting the file FILE 2: in this file I can see that the cloud of each sample is pretty narrow and there is never overlap in the clouds among samples.
> >>>
> >>> Now I am trying to interpret such results. Since in the bootstrapped tree I can see that certain samples cluster together with good support (for example MK4 and MK6 or MK11 and MK12),
> >>> I was expecting also the clouds of these couples of samples to overlap in the jackknifed 3D PCoA plot. But it is not the case.
> >>>
> >>> I would like to ask if I made everything right in the commands I used and if I understood correctly the meaning of these analyses, or if I making any mistake in the
> >>> interpretation of the results.
> >>>
> >>> Is it right to think that if the samples cluster together with good support in the UPGMA tree, they should also be "close" in the PCoA plot and with overlapping clouds?
> >>>
> >>> Thank you very much for any possible explanation and help.
> >>>
> >>> Best regards
> >>>
> >>> Niccolo'
> >>>
> >>>
> >>>
> >>>
>

> --
>
> ---
> You received this message because you are subscribed to the Google Groups "Qiime Forum" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to qiime-forum...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Sophie]

Hi Niccolo,

They are jackknifed values.  You *could* pass --subsample_multinomial to the rarefaction step and thus have actual bootstrap support (thanks Justin). But that rarely happens, and we don't recommend it.

Thanks,

Sophie

[Niccolo`]

Hi Sophie and Yoshiki,

thank you for the quick reply and for the information you provided. Now everything is more clear.

All the best

Niccolo´

[Yoshiki Vázquez Baeza]

Hello Niccolo,

We made a small maintenance release (0.9.51) yesterday, this release includes updated compatibility with Numpy 1.9.x, meaning that you will no longer see the warning you previously reported.

https://github.com/biocore/emperor/releases/tag/0.9.51

You should be able to upgrade using pip or ask your system administrator to do it for you:

pip install emperor --upgrade

Thanks!

Yoshiki.