See also
Downstream analysis using QIIME
If
you're using QIIME, you might be interested in some of the newer algorithms in
USEARCH which are not supported by the QIIME scripts.
UPARSE generates
much more
accurate OTUs.
Expected error read quality
filtering is more effective at discarding bad reads. The QIIME quality
filter allows many reads with >3% errors, causing many spurious OTUs.
Bayesian paired read assembly calculates posterior
quality scores, further improving quality filtering for merged reads.
The SINTAX and
UTAX algorithms give more accurate taxonomy
predictions as shown by these
taxonomy benchmark
results.
The UNOISE algorithm
reconstructs correct biological sequences from noisy reads, resolving
differences as small as a single base. UNOISE is thus able to distinguish
species and strains that would be merged into a single 97% OTU:
The
UNCROSS algorithm detects
cross-talk in Illumina reads. At the time of
writing (Jan 2017), this is the only available method for detecting
cross-talk to the best of my knowledge.