OTU co-ocurrence question

473 views
Skip to first unread message

Blair

unread,
Jul 19, 2012, 8:33:46 PM7/19/12
to qiime...@googlegroups.com
Hello Qiime Forum Folk,

I'd like to ask for some help with determining otu co-occurrence.  We have a set of 90 samples (3 dietary groups, 30 individuals per group).  All are human faecal samples and we have around 3000 sequences/sample (FLX-titanium).  I've passed the samples through the QIIME initial pipeline and we have alpha and beta diversity data that shows similarities and differences between the sample groups.  Next we'd be interested to find out if any species (otu's) are more likely to be found together, regardless of sample or other metadata.  The aim is to study cultured organisms in vitro, however the choice of species to study in co-culture is not trivial. We've got some ideas just by eyeballing the data but I noticed that the 'otu_category_significance.py' script should allow me to carry out co-occurrence analysis.  I've looked through the script details and generated a couple of output files, but am confused as to how to read the results.
The script information states "The script can also be used to measure co-occurrence. For instance it can also be used with presence/absence or abundance data for a phylogenetic group (such as that determined with quantitative PCR) to determine if any OTUs co-occur with a taxon of interest, using the ANOVA, G test of Independence, or correlation." 
So I've got an otu table, with abundance data and taxonomy information attached.  How to I look for pairs of otu's that might be linked more frequently that expected by chance?
I was also thinking of trying the "make_otu_network.py" script.  Again, whilst I can generate output files and look at these in Cytoscape I can' for the life of me work out how to look at otu - otu linkages (rather than sample - otu linkages).

My apologies if I'm missing something obvious (ie: being a bit dim).  Any help would be much appreciated.

I'm currently running QIIME 1.4.0 (macqiime on a macbook pro laptop).  print_qiime_config.py -t output is pasted below.

Many thanks,

Blair

Stef

unread,
May 31, 2013, 5:44:58 AM5/31/13
to qiime...@googlegroups.com
 Hi,
I found this old post while looking for help on co-occurrence analysis using 'otu_category_significance.py'. I have basically the same question/problem as Blair and would be very happy if anyone here could help!
Thanks so much!
Kind regards,
Stef

Stef

unread,
May 31, 2013, 5:46:02 AM5/31/13
to qiime...@googlegroups.com
PS: Should maybe mention that I am using Qiime 1.7.0 on aws...
Cheers!

Will Van Treuren

unread,
May 31, 2013, 1:32:01 PM5/31/13
to qiime...@googlegroups.com
Hi Stef, 

The way we would currently suggest doing co-occurrence analysis is using one of several programs not included with QIIME. They perform better than what we have (though we are working hard on improving things) because they take in to account some of the important considerations when working with 16s survey data (compositionality and sparsity). I would suggest using one of the tools outlined in this post

As far as 'make_otu_network.py' is concerned, I think you will find that script inapplicable for your task. It makes a bipartite network only, so OTU nodes are connected to Sample nodes, but there are no connections directly between two OTU nodes or two Sample nodes. 

Hope this helps,

Will 
Reply all
Reply to author
Forward
0 new messages