I had several generations with energy 3 and something but they didn't appear to be the best ones. The minimal RMSD I have is 12.55.
I today recieved new PC from local PC manufacturer.
Downloaded the latest text client and set it up running as a service.
Checked the progress.txt after couple hours and got surprised...
http://www.matti.ee/~veix/wow.JPG
I looked progress.txt of my main AMD XP 2,3Ghz rig and energy values from 15-40 but nothing like that.
The PC is:
Pentium 4 HT 2,6GHz
Intel D865PERL
2x256MB DDR400
... and it has been rock solid running folding client.
I had several generations with energy 3 and something but they didn't appear to be the best ones. The minimal RMSD I have is 12.55.
allthough 4.752 is my current best energy... nothing appeared on the list too
maybe they think we're cheating unless there's another calculation method?
Energy reported in the progress.txt is not the same as the RMS shown on the stats - there are several posts about this but basically (iirc) as we are now Folding blind (i.e. we don't know the protein) we can't work out the RMS locally...I've seen an energy value (for this protein) of around -0.296 but I wouldn't expect to get an RMS of that...
If this method of finding the way that a protein folds was being used for real, the only way to check to see how good a fold is would be to check its energy. Proteins fold into a formation that is a low energy state. This version of the algorithm is checking to see how good a fold is by looking at its energy, it keeps the one with the lowest energy.
After it is submitted, the server checks the fold against what the actual protein looks like and calculates the RMSD of the fold, which is how different the fold is from the actual protein. Then the best 10 are displayed on the stats page. So the algorithm is not cheating by comparing its output directly to the actual protein, like it was on some other runs.
So the number that shows up in progress.txt and the stats are two different things. You can even have the same energy for two different folds and have very different RMSD values.
Thanx, that explains quite a lot
But then it's not possible to know RMSD locally ?
that is what they are sayingOriginally posted by [veix]
But then it's not possible to know RMSD locally ?
Use the right tool for the right job!