MED Issues with V4

[Ford Fishman]

Hi Meren,

I've been having trouble with MED removing many reads during decomposition.  In a dataset with over 7 million reads placed into hundreds of thousands of 97% OTUs by mothur, MED returned only 300 nodes under default settings.  When I decreased -M to 100, MED returned approximately 4500 nodes, still orders of magnitude lower than the number of 97% OTUs. 

Now, rare OTUs were not removed from the initial dataset, so this discrepancy could be caused by MED removing rare sequences.  Using a different dataset with ~2000 97% OTUs with rare OTUs removed, however, when I ran the dataset with all 2000 OTUs with -M set to 100, only ~900 nodes were returned.  

I'm wondering what exactly is happening here to cause this discrepancy between the number of 97% OTUs and nodes and if you would have any insight on this issue.  

Thanks,

Ford