Denoising pyrosequencing reads by flowgram clustering
This page hosts the software described in our paper
We understand our program as a fast, heuristic alternative to Chris
software suite (now called Amplicon noise). Chris introduced the
concept of flowgram clustering and he deserves all the merits. If
you are interested in this topic you might want to have a look at his
Noise From Pyrosequenced Amplicons BMC Bioinformatics 12:38.
denoising pyrosequencing amplicon reads by exploiting
published in Nature Methods, Volume: 7, Pages: 668-669 (2010)
The Denoiser has been fully integrated into the qiime 1.3 release and all new releases will only occur as part of qiime.
Please check the Qiime website for the newest version.
The last stand-alone stable release of our software can be downloaded here:
This release contains updated error profiles for the
Titanium sequencing platform using data published by S. Balzer
and K. Malde
2010 Bioinformatics paper. We tested the new profile on one
of the Mock communities the we used for our original publication
sequenced on a Titanium machine. Our main findings from that
As a side effect, the new version should usually run faster for
most Titanium runs, since the runtime depends directly on
the number of clusters.
- Denoising Titanium data with the older FLX profiles is
less effective than denoising FLX data.
- Denoising with the new error
profiles and slightly higher clustering thresholds (both can
be simply set using the new --titanium switch) reduces the
number of false OTUs considerably.
- Even after denoising, Titanium data seems to have more
noise (and thus more false OTUs) than FLX data. Therefore,
we suggest to use a very aggressive
quality filtering (e.g. tight length bounds) prior to
denosing to remove as many suspicious reads as possible.
As a consequence, you should be very careful when directly
comparing Titanium and FLX data, denoised or not.
See the included INSTALL file for installation help and the README for a quick tutorial. Also check the FAQ for some frequent problems.
Compatibility with QIIME:
The 0.91 release is compatible with the current stable Qiime release (1.2.1).
However, in order to use the titanium switch from QIIME's denoise.py wrapper script you will have to update to the current QIIME SVN version or wait for the next QIIME release.
Integration into Qiime. No more changes here.
0.91: Provide a titanium hook for compatibility with QIIME
0.9: This is the first release that we adapted and
tested for Titanium data.
- Add Titanium error profile and --titanium command
line switch to change thresholds and use new error profile.
- Combine individual sample results after denoising samples separately
- add switch -m/--max_num_rounds to abort phase II early.
- add worst-case estimate of remaining rounds to log output.
- Fixed bug that left process hanging in rare cases.
0.851: Fixed compatibility issue with Qiime in
0.85: Replaced slow, file based client-server communication with sockets, which are a lot faster on most cluster environments. Add option to denoise all libraries/samples individually. Added FAQ.
0.84: Fixed an issue with how centroids for a cluster are chosen. Output fasta file now sorted by length, which allows e.g. uclust to choose better centroids with the --user_sort option. This version is consistent with the latest QIIME svn version.
0.83: Make scripts compatible with QIIME workflow scripts (pick_otus_through_otu_table.py). Compatible with the QIIME 1.0 release.
0.81: Added new script make_cluster_jobs.py for submitting to qsub based clusters running PBS.
0.8: Initial pre-release version
For questions or comments, please contact
jens.reeder(at)gmail.com or visit
the QIIME forum
Last edited March 9th 2011.