Having asked for secondary structure predictions and been given them, I wonder if it's such a good idea after all :) On looking at the Freestyle Mini-CASP (186) my immediate thought was to start with the sheets and join them up in an anti-parallel way so they all lay side by side and proceed from there. Judging by some of the other structures I saw, I'm guessing that others had similar ideas. Yet this approach yielded structures that scored 500 points or so less than the slightly ugly-looking 'furball' structures (188-190) we're seeing as starting points generated by Rosetta.
I'm slightly surprised that secondary structure prediction isn't quite reliable. Intuitively you'd have thought getting this right would be a prerequisite for determining the tertiary structure.
Yes, this is one of the harsh lessons in protein structure prediction and why we were reluctant to assign secondary structure.
One of the main problems is that most secondary structure predictions return how confident they are with a certain secondary structure element, but this is hard to translate in the game.
For example, the SAM-T08 server (http://compbio.soe.ucsc.edu/HMM-apps/HMM-applications.html) reports these cool sequence logos (I have attached the sequence logo for SAM_T08's mini-CASP2 secondary structure prediction) where the red H is for helix, green E for sheets, and C for loops (coils).
In these sequence logos, the higher the letter is for a given amino acid, the more confident the prediction is.
For example, at residue 60, the F (Phenylalanine) is strongly predicted to be in a helix (and indeed all three Rosetta@Home solutions and the freestyle puzzle have a helix for Phenylalanine 60).
But the P (Proline) at residue 31 basically has an equal prediction of being in a sheet or a loop (the green E and grey C have about the same height) so all we can take from this is that it is most likely not a helix.
It could be interesting to incorporate these different confidences into the game (coloring secondary structure elements by how confident the predictions are) but the game might get even busier than it already is!
You could post the sequence logos in a comment at the puzzle's page.
By the way, there is no attached sequence logo within your posting. Maybe this is caused by the bug I reported earlier: http://fold.it/portal/node/986615
Great idea, Madde… it seems like nobody is using the puzzle comments, so this might be a good use for it!
Can you now see the attachment in my previous post?
Yes, now I can see it.
Thank you!
