OTU picking

[beurmann]

Aloha,

I would like to get your input on how and when in the analysis process you pick your OTUs. Let's say you have 14 different samples and you have 5000 HQFL reads per sample. Would you run the OTU picking for each of the 14 samples individually, or do you put all the reads from the 14 samples together and then pick the OTUs? To me the first option sounds more appropriate, but I know some people pick their OTUs before they separate the reads by sample. I mainly wonder if you introduce any kind of bias if you do it one way or the other and how it affects the downstream analysis (i.e. beta diversity analysis)...?

Mahalo

[cliffbeall]

I think you should pick OTUs on the complete set of sequences, otherwise it seems like you could incorrectly define essentially the same OTU in different samples as different, introducing errors (e.g. for calculating Bray-Curtis dissimilarity).