M5.32 - Tests and validation of the services in coordination with Workpackage 3

Report from WP3

WP3 asked various real users of phylogenetics applications to test out the services offered by OBOE and to give their free assessment. Some of the returns are copied below. WRT comments about the versions of Mr Bayes this has been updated on OBOE.

...............

Dear all,

a long time ago, Thierry asked about the advantages/distadvantages of the OBOE (Oxford) web server. While Gontran made / is making some practical tests using OBOE, here I want to outline some general points, pros and cons:

There is definitely a need for a good web server performing phylogenetic analysis. So far, the "Bioportal" in Oslo was the main webserver in Europe offering free computing capacity. Unfortunately, the Norvegian government does not want to subsidize non-Norvegian activities anymore. Consequently, unless you have a Norvegian collaborator, you will not be able to use this server anymore...
The other "big player" is of course the US American CIPRES gateway, which you can still use after free registration. In the past, I have been using this server capacity regularly, so I can make some comparisons between OBOE vs. CIPRES.

1. The handling of both surfaces is rather easy. In OBOE, you only have to upload your files, in CIPRES you have more choices (depending on the software you want to run) to upload files & configuration files or make your setting online. At CIPRES I like to option very much that you can follow your analyses almost 'real-time'! Then you see whether you made any mistakes or your analyses take too long etc. CIPRES have some browser incompatibility issues (especially with Google Chrome - so simply don't use Chrome!!) which is not the case for OBOE.

2. Currently, the possibilities of OBOE are very restricted. You can only select between BEAST and MrBayes - both are performing Bayesian inference of phylogeny... I do miss other methods such as likelihood analyses (since they are even more time-demanding!!). Now, the "old" MrBayes version 3.1.2 is available at OBOE, so you cannot benefit from the novelties of MrBayes 3.2.1 (such as improved parameter settings, clock-analyses etc.). Using CIPRES, you have BEAST and the new MrBayes (3.2.1) as well as several further programs. RAxML (and GARLI) are definitely very important for maximum likelihood analyses, and they are quite speedy on CIPRES. If you want, you can still use PAUP* (for example for parsimony anlysis) on CIPRES.

3. Regarding multiple sequence alignment, Muscle is installed on OBOE, while a number of other programs (incl. Muscle) is running on CIPRES. Nevertheless, I prefer to make my alignments with MAFFT using their Japanese web server.

In summary, I give a strong preference to CIPRES (especially due to the possibility of ML analyses).
Sorry for being so late, maybe this information is not completely useless...

Cheers,
Dr. Zoltan Tamas Nagy
Brussels 16 September 2013

...............

Hello Thierry,
Yes partly. Here are some observations:

Positive aspects:
+ Very easy access: login, listing of all analyses with their status (finished, running), e-mail informing when analysis is complete, easy download of results etc.
+ NO BUGS! I made the same analyses with another platform where I encountered plenty of bugs.
+ Validation ok: results obtained with Muscle and MrBayes are identical to those obtained after running these programs locally with the same programs.
+ Very useful to be able to run several analyses in parallel.
Negative aspects:
- As Zoltan already noted, OBOE proposes the old version of MrBayes (3.1). It would be nice to be able to run MrBayes 3.2
- Difficult to evaluate the speed since the other platform that I tested always ended up with a bug. As expected, analyses were much faster than when they were performed on a personnel computer. However this was not always true with some smaller datasets. This probably depends on the waiting queue.

Suggestions:
1) Is it possible to sort the listing of all analyses (under "my analyses") or to make it searchable?
2) Is it possible to add a basic quality control of the submitted files in order to check that it is executable by the program available on OBOE (for example by MrBayes). Another possibility would be to advice the user to test his file locally on his personal computer before submitting it to OBOE. This might help the server to exclude some jobs instantaneously.
2) Would it be possible to give access to a server on LINUX in order to perform NGS analyses? This would be great for JEMU (no idea if this is possible with the current system).

Conclusion:
I find OBOE very useful (I used it several times already). Hence it would be great to have access to updated versions of the currently available programs and to other programs for other kinds of analyses (e.g. Max Likelihood).

Yes China is getting closer and closer, in a few hours, I ll be in Bruxelles Midi :-)
Have A nice travel to Firenze!!
Cheers,
Gontran Sonet
Joint Experimental Molecular Unit (JEMU)
Royal Belgian Institute of Natural Sciences
29, Rue Vautierstraat, 1000 Brussels, Belgium
http://jemu.myspecies.info/

Brussels 24 October 2013