Sat Aug  2 13:35:14 PDT 2008
TR469

Sat Aug  2 13:39:59 PDT 2008 Kevin Karplus

CASP organizers say

    REFINEMENT TARGET TR469 (the best submitted model according to
    GDT_TS). GDT_TS=82; RMSD=1.84A. First two N-term residues trimmed
    in the exper. structure and the model  

Make started Sat Aug  2 13:41:18 PDT 2008
Running on moai13.kilokluster.ucsc.edu

Sat Aug  2 18:51:02 PDT 2008 Kevin Karplus

The model to refine is Zhang-Server_TS3.
My closest model is try10-opt1 at GDT 96%

My closest submitted model is try13-opt3 (my model 1 at 94.8%).

Since the Zhang3 model has GDT_TS=82%, my models are much closer to
zhang3 than it is to the correct model.  By comparing models, I may be
abel to figure out which way to go.  Do I want to continue trying to
cluster the CCH?


Mon Aug  4 11:50:11 PDT 2008 Kevin Karplus

The models are all quite close.  I think I do want to add the CCH constraints.

Mon Aug  4 14:00:06 PDT 2008 Kevin Karplus

try3, optimized with the force_HCC constraints does score better, but
mainly from reducing clashes---the HCC constraints are only slightly improved.

Tue Aug  5 12:05:41 PDT 2008 Kevin Karplus

try4 optimizes try3-opt3 and improves scores slightly, but rosetta
does not like it as much.

Tue Aug  5 14:01:20 PDT 2008 Kevin Karplus

For try5, I'll try optimizing from-try13-opt3, which has a slightly
different loop for G31-A38.  I think it  needs to e packed a bit tighter.

Tue Aug  5 14:09:50 PDT 2008 Kevin Karplus

The biggest differences between the models seems to be the packing of
the aromatics: F11, F5, and Y28.

Tue Aug  5 14:33:00 PDT 2008 Kevin Karplus

try5-opt3 has slight clashes and breaks that may need to be removed.
It may also be driven too much by the HCC constraints.

I don't like the position of Q3 in try5.

All my models look like they need to be squeezed a bit tighter.

Tue Aug  5 15:12:58 PDT 2008 Kevin Karplus

try6 squeezes try5 a bit, but introduces more clashes and breaks.
The scaling of soft_clashes in try6 is too fast, as each optimization
discards the previous one and goes back to try5-opt3.

Tue Aug  5 15:41:43 PDT 2008 Kevin Karplus

try7 gets a new best, both for undertaker scoring and for rosetta.

But in try7, the C32 and C35 SG atoms are a bit too close at 2.56 Ang,
instead of 3.6  For try8, I'll modify the distances to be more tightly
constrained around the desired distances.

Tue Aug  5 16:03:13 PDT 2008 Kevin Karplus

try8 started, but only using the gromacs-optimized models from try5,
try6, and try7, to try to weed out models with clashes.  I think that
gromacs may have stretched some bond distances to remove clashes,
though, which I'm not sure undertaker can recover from.

Tue Aug  5 18:03:50 PDT 2008 Kevin Karplus

try9, from try8, does not score quite as well as try7 with the try9
costfcn, and rosetta prefers try8, then try7, then try9.

For try10, I'll use the try9 costfcn, but try optimizing from tries 2,3,4.

After that, I'll probalby make a chimera, taking the best aspects of
each model that I can find.

Tue Aug  5 21:32:02 PDT 2008 Kevin Karplus

try10-opt3 scores almost as well as try7-opt3, though rosetta doesn't
like it that much.

I made a chimera-try10-try7 which is mostly try7-opt3, but takes Q3
and K4 from try10-opt3.  It has a few clashes, but can probably be
improved.  I'll try optimizing it in try11.

Tue Aug  5 22:15:33 PDT 2008 Kevin Karplus

try11-opt3 looks pretty good to me (and to the try11 costfcn).
There may be a bit more tweaking possible, but I think this is about
as good as it gets.

Tue Aug  5 22:52:47 PDT 2008 Kevin Karplus

try12 made tiny improvements in try11, and rosetta almost likes the
repacked version as well as it likes try8.


Thu Aug 14 14:39:48 PDT 2008 Kevin Karplus

I still don't see anything to do, so will submit


ReadConformPDB TR469.try12-opt3.pdb # < try11-opt3 < chimera-try10-try7

ReadConformPDB TR469.try10-opt3.pdb # < try4-opt3 < try3-opt3 < try2-opt3 < TR469

ReadConformPDB TR469.try9-opt3.pdb # < try8-opt3  < try7-opt3.gromacs0 < try6-opt3 < try5-opt3 < from-try13-opt3.pdb< try6-opt3 < try5-opt3 < from-try13-opt3.pdb

ReadConformPDB TR469.try8-opt3.gromacs0.repack-nonPC.pdb # < try7-opt3.gromacs0 < try6-opt3 < try5-opt3 < from-try13-opt3.pdb
	# best rosetta energy

ReadConformPDB TR469.try4-opt3.pdb # < try3-opt3 < try2-opt3 < TR469