We just finished teaching a second version of our two-day shotgun metagenome analysis workshop, this time at UC Santa Cruz (the first one was in October 2016, at Scripps Institute of Oceanography). Harriet Alexander led the workshop and Phillip Brooks and I co-taught; Luiz Irber, Shannon Joslin, and Taylor Reiter TAed. The workshop was hosted by Professor Marilou Sison-Mangus at the Earth and Marine Sciences Building.
(Note that Harriet will be running an expanded version of this workshop at our summer institute, July 17-21. Registration is still open!)
About 30-35 people came the first day, and about 30 were there on the second.
Some good - new lessons!
- binning genomes out of your metagenomes with MaxBin - we used MaxBin, which was fast enough to use in an hour lesson.
- Quickly searching and comparing your samples with sourmash - this is the first time we've taught sourmash to anyone outside the lab, and I think it went pretty well. Plus we now have a sourmash tutorial!
More good - using XSEDE Jetstream instead of Amazon Web Services!
This was the first genomics workshop in many years where we didn't use Amazon Web Services - we used XSEDE Jetstream instead. See our login instructions here.
Why are we abandoning Amazon? Two reasons --
- while we've been teaching it for almost 8 years now, the conversion rate seems to be very low: AFAICT our students aren't using it, because it costs money and their advisors don't want to pay for AWS when they can use institutional resources. (This is anecdotal.)
- since sometime before October 2016, Amazon changed their registration system so that newly registered people cannot start up instances for a few hours after their first try. This is death on half-day and two-days workshops. (You can read a bit more about it here.) There seems to be nothing that AWS folk can do to help us so we are giving up.
I am happy to report that Jetstream went more smoothly than AWS in almost every way and seems to perfectly meet our needs for training! We may have more to say about it after our summer institute's use.
I also suspect that people will be more inclined to use Jetstream if they can get allocations on it for free; there was significant interest in this during the workshop.
Other good --
- As always, the people that attended the workshop were fantastic, and dealt with our occasional hiccups pretty well!
- We managed to pretty smoothly move between the command line and the Jupyter Notebook for two of the lessons, which was pretty cool.
- We managed to implement a simple demo of a tetramer nucleotide frequency clustering system using sourmash and t-SNE - see the notebook on github (which should be run after the initial steps in the binning lesson).
There is no bad or ugly
Nothing went wrong! Which I guess is a 'good' all on its own!
There were a few minor issues with the Jetstream desktop, and some problems with starting up Jetstream instances every now and then, and the guest network at UCSC blocked port 8000 (which we used for Jupyter), but most of the time we could work around these issues.
Feedback from participants
The in-person feedback (which is admittedly always kinder than the anonymous feedback :) was excellent - students really liked the hands-on teaching style (Carpentry-style, but with copy/paste) and the slow teaching pace with lots of time for questions was well received.
As always, our materials are available under CC0 on github - the URL is https://github.com/ngs-docs/2017-ucsc-metagenomics.