Vsearch cluster results seems inconsistent

[Hu Peiyao]

Hi,

When I try to reconduct 16S amplicon analysis that had been analyzed by my colleague before with vsearch (some parameters maybe different), the cluster results is different. Some CoreOTU(higher occurrence among samples) are different from my colleagues, How to configure the precise cluster results, or how can I optimize the pipeline?

I'm glad to hear your suggestions.

Any help will be GREATLY appreciated!

[Frédéric Mahé]

Hello,

clustering results can vary for obvious reasons:

- different command-line parameters,

- different vsearch version (some older versions might yield slightly different clustering results; not very likely but possible)

Also, threshold-based (e.g.; 97% similarity) clustering results are sensitive to data input-order. If fasta entries are the same, but in a different order, clustering results might be different.

Finally, in some cases, multi-threading may change the output order of clustering results (but not the content of each cluster, if I am not mistaken).