"horseshoe effect" in PCoA based on Unifrac

[Becks]

Hello!

I have a beautiful "horseshoe effect" due to a gradient in my
bacterial data set that is quite noticeable in the PCoA generated by
beta_diversity_through_plots.py. The PCoA is calculated from the
unweighted unifrac distance matrix.

I need to try and remove this from the data with a transformation, but
not sure how to deal with this in QIIME? Or if there is a "best"
transformation for the OTU table when using Unifrac?

Thanks for any advice!
Becks

[justink]

I don't think most OTU abundance transformations are going to help with a horseshoe in unweighted unifrac, as unifrac's using only the presence or absence of an OTU, and thus e.g. taking the sqrt of the abundance isn't going to change the unifrac result. There's definitely some research into beta diversity methods which don't show horseshoe effects, but I'm afraid none of those were mature enough to warrant inclusion in qiime, and we specifically avoided detrended correspondence analysis. Sorry I don't have a better solution for you.

[Rebecca Prescott]

Hello!

Thanks for your input.  I realized the same regarding the usual transformations after I posted because I was dealing with presence/absence data.  My concern is that I want to run a db-RDA, which I understand is based on the PCoA analysis.  I am concerned that if the gradient is present it will influence the db-RDA?    Any advice on how to proceed would be great.  I have thought about using nmds instead, but not sure what I could use that would be similar to a db-RDA to follow?  I am trying to determine what treatments are driving the changes in biofilm communities.  

Dr. Bowers and I were working on this yesterday, and I found a new QIIME command - detrend.py in the latest version of QIIME that is suppose to address this.  Does not appear to be in previous versions of QIIME.  I think its based on the attached paper where they applied a quadratic function to the ordinations to try and remove them.  Usually, you would address such issues with transformations on the original data set, so we were wondering what your opinion might be on their method?  Also, Bob couldn't get the command to run in MacQIIME - we kept getting an error message that the data "were not iterable."  We tried inputing the coordinates file, the distance matrix and the OTU table - just to see if we could get any of them to work.  We also tried another dataset that he had with a similar PCoA and got the same error message.  

Becks

beachb...@gmail.com

[Jai Ram Rideout]

Hi Rebecca,

I'm not sure if db-RDA will avoid the horseshoe effect since it is a constrained version of PCoA, but you might give it a shot. QIIME offers some limited db-RDA functionality in compare_categories.py, and we have a tutorial here:

http://qiime.org/tutorials/category_comparison.html

Hope this helps,

Jai 

--

---
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/groups/opt_out.

--
 
---
You received this message because you are subscribed to a topic in the Google Groups "Qiime Forum" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/qiime-forum/BbzAQrBDxmw/unsubscribe?hl=en-US.
To unsubscribe from this group and all its topics, send an email to qiime-forum...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Dan Knights]

Hi Becks,

The quadratic detrending approach is useful for visualizing data that run along a gradient; using it to fit linear models like RDA has not been tested.

Can you please supply the command and input files that you used for detrend.py? You can send them directly to me if needed.

Thanks,

Dan