I am analyzing 16S rRNA gene survey (V4 region) of human fecal microbiota. During DNA extraction with MoBio PowerSOIL DNA extraction kit I've left one sample without the DNA for each extraction I've performed. This "blank" was used in order to account for "kitome" - see article here.
PD. Schloss (author of this article as well as "mothur") has not at that time suggested any particular statistical/bioinformatic technique to account for "contaminant" microbial abundances.
One approach I've used up until now -> to use blast searches (qiime's exlucde_seqs_by_blast.py) and discard sequences that match in 97% or 98%. This threshold is somewhat unclear. However, there is a huge drawback to this approach. One can imagine a certain bacteria present both in studied samples (as fecal microbiome of patients) and the same bacteria (but with different abundance) present in "extraction blanks". With this approach - blasting and removing let's say 98% similarity matches - I'll just remove this bacterial sequences from the analysis.
I don't think this approach is therefore valid.
I am thinking now about SUBTRACTING counts from OTU tables. I.e. if there is certain bug present in both samples (studied and dna extract control) then counts from control should be subtracted from the studied sample within the same OTU.
What do you think? It this more valid? Are there any other methods that could help me with "kitome"?
Best regards, Robert.