I’ve recently analyzed the microbial composition of some gastrointestinal biopsies in different disease states with mothur (454 sequencing). We had sampled 4 sites throughout the gastrointestinal tract. I have some really perplexing results. 2 patients with the same disease (in all 8 sampled sites) have a count of zero bacteroidetes… This is quite strange as all of my other samples have a very large count for obvious reasons (being sampled from the gut).
Do you have any idea if this (not-so-right-results) could be a result from the mothur analysis, or perhaps an error in 454 sequencing?
Hmmm, that is weird and it’s great that you’re thinking critically about such things.
A couple of thoughts…
Were all of your samples processed in parallel? Did you use dereplicate=T in chimera.uchime (see the SOP)? If not, it’s possible that a rare Bacteroidete got flagged in one sample, but was uber abundant in the ones that now don’t appear to have any.
Were the subjects treated differently? I could imagine an antibiotic treatment might wipe them out of some patients but not others.
Sorry, could you clarify what you mean by processed in parallel? They should have been sequenced in the same 454 run, and all samples were analyzed with mothur together. These samples were also analyzed a few years ago with an older version of mothur by someone else and there was the same issue (actually, this is why I began re-analyzing this data). So at that time the dereplicate parameter wasn’t an option. And in this re-analysis I also did not set dereplicate to T. I could try that though! Do you have any other suggestions of what I might try?
Subjects were actually screened for antibiotic treatment so there would be no way an antibiotic would influence these results.
Try it with dereplicate=T. Also, you could try to just run the raw sequences through classify.seqs and see if those samples have Bacteroidetes before any of the error correction steps.