Loadings for PCoA?

[MBA]

Hi,

Is there any way to view the loadings on each principal component
generated by the principal_coordinates.py script? I can successfully
run the script, and then make 3D and 2D plots... it would be helpful
to know which OTUs (or taxa) are loading each component.

Thanks!
M

[Jeff Werner]

Hi Meghan,

Because the principal coordinates are based on a distance matrix, there aren't really otu components that you can graph. I think you would have to do principal components aanalysis of the otu table to get the loading of individual components. That might be somethiing you would have to do in a different software package...

You can use taxonomy summary data in the mapping file to color in the pcoa plots, to visualize trends in taxable abundances, but that is about it for principal_coordinates.py.

Cheers,
Jeff

[Luke Ursell]

Hi,

I would also recommend running otu_category_significance.py. This will allow you to determine which OTU's are significantly different between samples (based on metadata classifications), and might provide some insights into the OTU's that are separating out groups on your PCOA plots.

Best,
Luke

--
Luke Ursell
Graduate Student - Knight Lab
University of Colorado Boulder

[Meghan Azad]

Thanks Luke & Jeff for your helpful responses!

On Nov 17, 2:44 pm, Luke Ursell <luke.urs...@colorado.edu> wrote:
> Hi,
>
> I would also recommend running otu_category_significance.py. This will
> allow you to determine which OTU's are significantly different between
> samples (based on metadata classifications), and might provide some
> insights into the OTU's that are separating out groups on your PCOA plots.
>
> Best,
> Luke
>

> On Thu, Nov 17, 2011 at 10:27 AM, MBA <azad.meg...@gmail.com> wrote:
> > Hi,
>
> > Is there any way to view the loadings on each principal component
> > generated by the principal_coordinates.py script?  I can successfully
> > run the script, and then make 3D and 2D plots... it would be helpful
> > to know which OTUs (or taxa) are loading each component.
>
> > Thanks!
> > M
>
> --
> Luke Ursell
> Graduate Student - Knight Lab
> University of Colorado Boulder
>

> <Luke.Urs...@colorado.edu>

[justink]

One thing I'd like to note - although PCoA isn't amenable to axis
loadings in the same way PCA is, you can correlate taxa to the PCoA
axes, and QIIME allows this.

The process is pretty much to use summarize_taxa.py followed by
make_3d_plots.py -t ... There's a bit of info on the documentation for
make_3d_plots at the bottom of the page:

http://qiime.org/svn_documentation/scripts/make_3d_plots.html

[MikeyJ]

That's very useful - thanks,

Does the size of the area shaded around the taxa point indicate the
strength of the correlation or abundance of that taxa?

Cheers

Mike

[Meghan Azad]

Thanks! I'll give this a try.

On Nov 23, 2:34 pm, justink <justi...@gmail.com> wrote:
> One thing I'd like to note - although PCoA isn't amenable to axis
> loadings in the same way PCA is, you can correlate taxa to the PCoA
> axes, and QIIME allows this.
>
> The process is pretty much to use summarize_taxa.py followed by
> make_3d_plots.py -t ... There's a bit of info on the documentation for
> make_3d_plots at the bottom of the page:
>
> http://qiime.org/svn_documentation/scripts/make_3d_plots.html
>