Hello group, I am posting here for the first time.
We developed a new secondary structure assignment program (SST) based on the Bayesian method of minimum message length inference: http://bioinformatics.oxfordjournals.org/content/28/12/i97.abstract?keytype=ref&ijkey=AWyPEpQZaKi7Hne Minimum message length inference is a general framework from the statistical learning literature which links hypothesis/model selection with information theory and data compression. SST treats assignments of secondary structure as hypotheses that attempt to explain the observed protein coordinate data economically/concisely. The core idea is that the * best *secondary structure assignment should be able to explain (that is, losslessly compress) the (CA) coordinates in the most concise way. This is equivalent to maximizing the joint probability of a hypothesis and the data, thus giving SST a good mathematical foundation to address this problem. SST attempts to: (1) delineate secondary structural elements -- helix and strands of sheet -- in a statistically consistent way. (2) detect pi- and 3_10- helical caps/segments and bulges, in addition to the standard alpha-helices (3) assemble the various strands into a consistent sheet (4) provide a readable output of dissected secondary structural elements, with a companion pymol-loadable script to visualize various dissected secondary structure elements independently. A quick-and-dirty web server has been set up here in the interim: http://lcb.infotech.monash.edu.au/sstweb We plan to release a standalone version soon, but I would like to rectify any obvious problems before that. We will be grateful if the readers here can make suggestions to improve this program. Thanks, Arun Konagurthu Monash University
