See also
QIIME does not support state-of-the art algorithms
in USEARCH
Downstream analysis in QIIME
I am familiar with the basics of QIIME; for example, I
have run some of the scripts for benchmarking. But I am not an expert -- if any
of the information on this page is wrong or out of date, please let
me know and I will update it.
Running usearch in QIIME
Some QIIME scripts run old
versions of USEARCH(or the even older uclust binary). I am not clear which
versions of usearch / uclust are compatible with which versions of QIIME,
and I don't know how the installation works. For more information, please
refer to the QIIME documentation. As far as I know, usearch v6.1.544 is the
most recent version with any support in QIIME.
Recommended: Denoise or generate OTUs with USEARCH
UPARSE generates OTUs which are
far superior to any of the methods currently supported by QIIME (see
benchmark
results).
UNOISE recovers the biological sequences in a sample by "denoising" (error-correction), giving better resolution than 97% OTUs. The denoised sequences can be considered OTUs and can be used to make an OTU table (see unoise command for details). I recommend denoising over making 97% OTUs in most cases. QIIME does not support denoising, at least through v1.9.1.
Recommended: Predict taxonomy using UTAX or SINTAX
The
UTAX and SINTAX
algorithms give better prediction accuracy than any of the methods supported
by the QIIME assign_taxonomy.py.script, all of which have
high over-classification rates.
Downstream analysis
Once you have made an OTU table,
then you can use QIIME (or mothur) for downstream analysis: alpha and beta
diversity, principle component analysis etc. USEARCH can generate OTU tables
in QIIME classic tabbed
text format and
BIOM v1.0 format (JSON). The
biom
utility can be used to convert to
BIOM v2.1 format (HDF5). This should give you everything you need to do
the rest of your analysis in QIIME. If it doesn't, I will add new features
to USEARCH as needed --
just let
me know.