I have seen many comments at being able to analyze OTUs at different “taxonomic” levels.
Qhestion - isn’t that what Metastats is designed to do? Could we not simply create design groups based on the OTUs and analyze them? The feature being the different taxonomic levels assigned to the OTUs? Still trying to think this one through - comments would be appreciated.
I guess what I was wondering, is there a way to set up the MetaStats program to accept a taxonomy summary file as input? This would allow for testing across samples at different levels of taxonomic assignments across all our different groups. And would be more applicable towards the original intent of the creators of that program - correct?
metastats is a non-parametric T-test. So you can give it any shared file and design file and it will give you the P-values and corrected q-values to tell you the significance. You could do this based on OTUs or phylotypes. Perhaps I’m not following what you’re asking?
If I understand, there are two “types” of metastat programs.
The “Metastats: Detect Differentially Abundant Features” portion of the program could be used to look at different assigned taxonomy levels, correct?http://metastats.cbcb.umd.edu/detection.html.
The metastats paper has examples of this using different levels of assigned taxonomy and functional assignments for genes (Tables 2 and Tables 4 in White et al (2009) Statistical Methods for Detecting Differentially Abundant Features in Clinical Metagenomic Samples Plos One, http://www.ploscompbiol.org/article/info%3Adoi%2F10.1371%2Fjournal.pcbi.1000352
All it needs is a list of features and the number of times they occur in the samples. The taxonomy summary file is in a “correct” format to use for the analysis - provided that the metastats program incorporated into mothur does a normalization procedure as the program. Otherwise, a rabund file of the taxonomy would be sufficient.
Their webserver (the address above) has a version available, but currently it is not functioning (there is a bug).
So, what kind of shared file would I generate to get something that resembles the taxonomy summary file? I may have misunderstood what you misunderstood. I am sure you know what metastats is designed to do , I am “thinking out loud” to figure out if I can apply this to my dataset.
So have you tried using the output from make.shared? You can use phylotype to assign sequences to different taxonomic levels and get the counts for each taxa in all of the samples at different taxonomic levels. Then you could use this as input to metastats.
I am having difficulty understanding what the difference is between the phylotype command and the classify.otus command and how the output differs in terms of the actual OTUs and the assigned taxonomy… As I wanted to analyze the OTUs - and the taxonomy assigned to them, that is what was trying to do. I know you have explained it several times, but I guess can’t grasp if or what the distinction is between the two sets of results.
No, I have not done the phylotype command. Will try that. Thank you.
So if you want OTU classifications then you need to run classify.otu and this will give you the consensus taxonomy for each OTU. phylotype is analogous to cluster except that it is based on the taxonomy of your sequences. To figure out the actual taxonomy assignment of the clusters you will need to run classify.otu. You mentioned before that you wanted to run metastats at different taxonomy levels. The only way to do this and have it make sense would be to do phylotype and then run metastats on the different taxonomic levels (1=genus, 6=kingdom).