Tuesday, 23 April 2013

Protein function annotation

In the recent CAFA (Critical assessment of protein function annotation) paper, they compared the ability of 54 different methods to computationally predict the functions of 866 proteins from 11 different organisms.

I checked to see whether the software for any of the top-performing methods is available.

The top 6 methods in predicting 'molecular function' GO terms were:
1) Jones-UCL - this method is not available yet. I emailed the author, and he said that part of it is available as the FFPRED web server, which uses a feature-based approach to function prediction, but doesn't use the orthology and homology components of the Jones-UCL method used for CAFA. There is a paper by Cozzetto et al 2013 on the Jones-UCL CAFA method.
2) Argot2 - there is a web server available, but you can only submit 5000 sequences at once. The program is not available for download. There is also a paper by Falda et al 2012.
3) PANNZER - there is a website, and should be soon a paper, a web server and program available for download but it's not there yet.
4) ESG - there is a web server, but you can only submit 10 sequences at once. There is a paper by Chitale et al 2009.
5) BAR+  - there is a web server, but you can only submit 50 sequences at once.  There are papers by Bartoli et al 2009, and Piovesan et al 2011.
6) PDCN (MULTICOM-PDCN) - there isn't a web server or software for download yet, but I emailed the authors and they told me that a web server will be available soon. There is a paper by Wang et al 2013.

These top-performing methods had F-measures (a measure of prediction accuracy, that can have a maximum of 1) of about 0.54-0.60, compared to about 0.4 for BLAST.

The top 6 methods in predicting 'biological process' GO terms were almost the same:
1) Jones-UCL
2) Argot2
5) ESG
6) Rost Lab - there is a paper by Hamp et al 2013. The software can be downloaded as the program 'Metastudent' from the website.

These methods had F-measures of about 0.37 to 0.4, compared to about 0.27 for BLAST.

Thanks to James Cotton and Adam Reid for bringing CAFA to my attention.

No comments: