Wed May 21 09:17:48 PDT 2008
T0411
Make started Wed May 21 09:26:14 PDT 2008
Running on cheep.cse.ucsc.edu

Wed May 21 10:40:33 PDT 2008 Kevin Karplus

PDB blast gets a good hit to 1qxnA, for most of sequence.

Wed May 21 11:34:16 PDT 2008 Kevin Karplus

HMMs are getting c.46.1.2 folds, while 1qxnA is c.46.1.3.
We may have to look more closely at the difference between these
families, to choose the appropriate templates.  There are lots of
c.46.1.* templates, so we may be able to discard some safely.

We have good agreement for S24 on, but the first 23 residues are not
placed consistently from the alignments.  The try1 run collapsed the
initial segment, and wound a predicted strand at M1-W9 into a helix.
The t06 predictions were mainly for a helix there, so I'm not sure why
the consensus ehl2 prediction was for a strand.

In some of the templates, the N-terminal region corresponds to a
linker between copies of the domain, so I wonder if I should try
making a dimer based on two copies.

The conserved residues, V44,D45,R47,I57,S60,C98,G104,G115,S125,W129,
cluster except for G115, which is at the other end of a helix.
D45,R47,S60,C98, and S125 probably form the active site of a sulfur
transferase.   Neither conserved residues nor rr predictions are much
help with the N-terminus.

I could try sticking on the strand that seems to be forming:

SheetConstraint N21	Y25	N118	Y122	Hbond T119

Fri May 23 05:22:32 PDT 2008 Kevin Karplus

Trying with the extra strand for try2.

Fri May 23 06:34:13 PDT 2008 Kevin Karplus

I think that try2 was working too hard on clashes and breaks, and not
enough on the extra strand.  I'll rebalance the weights for try3 and
try again.

Fri May 23 08:45:22 PDT 2008 Kevin Karplus

try3-opt3 gets some of the desired strand (perhaps as much as is feasible):
SheetConstraint (T0411)S120 (T0411)L121	(T0411)Q23 (T0411)S24	hbond (T0411)L121	1

but has worse clashes and breaks than try2 or try1.
Maybe I should start a polishing run from the try3-opt3 models, using
a costfcn that penalizes clashes and breaks more.

Wed May 28 11:15:36 PDT 2008   SAM-T08-MQAO hand QA T0411 Submitted
Wed May 28 11:15:36 PDT 2008   SAM-T08-MQAU hand QA T0411 Submitted
Wed May 28 11:15:36 PDT 2008   SAM-T08-MQAC hand QA T0411 Submitted

Sun Jun  1 11:33:08 PDT 2008 Kevin Karplus

The models pretty much agree after Q23, but differ considerably before that.
The Y65-A91 loop also varies, though try1-opt3 seems to be an outlier
here.

I'll do the two meta sever runs to see what comes out, then try a
polishing run using all the models.

Sun Jun  1 12:11:30 PDT 2008 Kevin Karplus

MQAU1 is based mainly on Phyre_de_novo_TS1, as is MQAC1.

Both score better than the models we built from alignments.
There are some small differences from Phyre_de_novo_TS1 at the
N-terminal end and in the packing of F27 and W129.

I like the N-terminus of MQAC1, but prefer the hydrophobic packing of MQAU1.

The biggest differences in the core between these models and our try1
is for S67-A91.  I may actually prefer try1-opt3 to MQAC1 in this region!

Let me  make a chimera:
	M1-Q23 	from MQAC1
	S67-A91 from try1-opt3.gromacs0
	rest 	from MQAU1

Sun Jun  1 13:08:24 PDT 2008 Kevin Karplus

Optimizing that chimera as try5.  Assuming that works, I'll want to
select models with different N-termini and different S67-A91 loops as
my models.

Sun Jun  1 14:07:50 PDT 2008 Kevin Karplus
 
try5-opt3 scores better than MQAU1 and almost as well as MQAC1.
I like the F27+W129 packing of try5-opt3 better, but I'm not so sure
about the S67-A91 loop---both seem decent, but MQAC1 scores better on
pred_alpha_back, pred_bys_back, and pred_pb_back.

Currently I'm favoring submitting

ReadConformPDB T0411.MQAC1-opt3.pdb
ReadConformPDB T0411.try5-opt3.pdb
ReadConformPDB T0411.try1-opt3.gromacs0.pdb
ReadConformPDB T0411.try4-opt3.pdb
ReadConformPDB T0411.try2-opt3.gromacs0.pdb

But I might want to do some more optimizing still.

Sat Jun  7 09:14:11 PDT 2008 Kevin Karplus

Let's see if I can improve the try5 run by upping clashes and breaks
and starting from just the gromacs optimized models (for all tries).

Sat Jun  7 10:41:15 PDT 2008 Kevin Karplus

try6-opt3 scores best with the try6 costfcn (finally beating MQAC1-opt3)
Rosetta likes try6-opt3.gromacs0.repack-nonPC best of all the models
it has repacked, but undertaker thinks that the breaks and clashes are
worse than try6-opt3.

I doubt that I'll see much more on this, so I'll submit

ReadConformPDB T0411.try6-opt3.pdb
ReadConformPDB T0411.MQAC1-opt3.pdb
ReadConformPDB T0411.try1-opt3.gromacs0.pdb
ReadConformPDB T0411.try4-opt3.pdb
ReadConformPDB T0411.try2-opt3.gromacs0.pdb

try6 < try5 < chimera-mqac1-try1-mqau1
		mqau1 < Phyre_de_novo_TS1
mqac1 < Phyre_de_novo_TS1
try1 fully automatic  < alignments
try4 < try3 < alignments
try2  < alignments