Scoring Functions

[Beebleman]

Howard,

Just a quick question about the scoring function. I noticed on the results section the lowest energy and smallest rotational gyration scores were quite a bit off from the smallest RMSD.

What (if you are in a position to disclose at this point) optimizations have you put into the current scoring functions based on the results from Phase Ia? (too soon for feedback from Ib I'm sure)

Just curious.

Dan :cool:

[Brian the Fist]

You are correct. I don't have timne to go into great deal right now, but briefly we are using a third energy function which is similar to the 'Bryant' one we used before but seems to work better. We also look at how compact the structures are and at how much secondary structure they have compared to that predicted (those helices and sheets (arrows) that you see in the results section are 'secondary structure' and help determine the fold of the proteins - their location can be predicted with some degree of accuracy from just the protein sequence). Thus we are combining several factors to attempt to pick out the best structures from the set we generate. We won;t know for sure if we picked the best though.

[Scott Jensen]

How are we doing so far on the scoring? As expected? Worse? Better? What does all those math formulas project to be the expected lowest score? Maybe we'll beat expectations this time too. :D

[Brian the Fist]

The lowest possible score as we have it now is 0. As for what we expect, it is hard to say but the current best scores of 40 or so are 'reasonable'. This number is not RMSD so it cannot be compared to RMSD in any way whatsoever. There is no guarantee that a lower score here means a lower RMSD (although ideally it would). We'll only know for sure after CASP is done how well we do ;)

[plaidfishes]

Howard: So if 40 was reasonable, then todays 1.64 is unreasonable? Insanely Great? Indicative of problems? Fantastically lucky?

In one of the other educational threads, there is a bunch of info on how the results for RMSDA (and presumably, psuedoenergy) were an extreme value distribution. There was also discussion elswhere about how the "Best" values are likely to appear as if the numbers just fell over a cliff.

With todays top ten all less than 7.77, does this suggest that we have fallen over that cliff? Or is the fact of them all being less than 8 suggest that we have not seen the extremes of this EVD and it can go down still further?

Just what does a 0.00 value for psuedo-energy actually mean? Is it equal to the mythical perfect structure? Is it simply the lowest value possible within the system? Is it an arbitrary number which you never ever expected to pass? At the current rate of dropping, we might actually achieve that 0.00 any day now.