Tue Jul 8 13:46:36 PDT 2008 TR432 Tue Jul 8 14:57:23 PDT 2008 Kevin Karplus The high-quality TR432 model is MUProt_TS5 The SAM-T08-server_TS1 model is way down around GDT 77%. We had predicted GDT of 80.4% for MUProt_TS5, and 69.2% for SAM-T08-server_TS1, so at least MQAC recognized that the SAM-T08 server did poorly. The others that it rated highly seem to agree to a large extent with TR432. The MQAU QA assessment was much poorer, thinking SAM-T08-server was good (predicted GDT 75.8% and MUProt_TS5 lower at 72%). Tue Jul 8 15:03:39 PDT 2008 Kevin Karplus What the CASP organizer say is REFINEMENT TARGET TR432 (one of the best submitted models). Quite accurate model: GDT_TS=91; RMSD=1.34. Concentrate on refining loop 85-92. There is another area of possible improvement. I doubt that we can improve much on a 1.34 Ang CA-rmsd model. Make started Tue Jul 8 15:05:23 PDT 2008 Running on moai30.kilokluster.ucsc.edu TR432 does not score all that well with the align constraints or most of the other cost fcns. There are no sheets. I'll probably have to extract constraints from the TR432 model to keep it from moving too much. Tue Jul 8 19:08:53 PDT 2008 Kevin Karplus The TR432 and try1 models are quite close. I'll probably try copying the interesting loop from try1, then polish the resulting model. I'll extract helix constraints from TR432.pdb and include those in the costfcn. Tue Jul 8 20:44:29 PDT 2008 Kevin Karplus I also included a disulfide bridge for C79-C101, since the cys came out too close for that to be an accident. try2 started on the moai cluster to optimize from the chimera that has mostly the initial model, but Y85-R97 from try1-opt3. Tue Jul 8 22:15:33 PDT 2008 Kevin Karplus The try2-opt3 model looks good, forming the disulfide, packing the helices slightly better without moving them much. I'll do another optimization run from the try2 models that have been run through gromacs, to try to reduce clashes and breaks further. I'll also do a try4, with the same costfcn, but starting from the initial TR432 model, and with BackRub and BigBackRub turned up initially. Tue Jul 8 23:48:05 PDT 2008 Kevin Karplus try4 improves on TR432, but isn't as good even as try2. I might want to look at trying the loop from T19 to T32 from try1 in the model also. Wed Jul 9 08:32:21 PDT 2008 Kevin Karplus Perhaps more importantly, in relieving clashes I think that the final helix has moved a little too far out, leaving small voids. How can I pack it tighter? L115.CD2 V68.CB 3 L115.CD2 V68.CG2 2.97 L115.CD2 V68.CG1 3.10 L115.CB I111.CG2 3.22 Hbond K69.NZ E124.OE2 C123.CB V68.CG2 from 6.56 down to 4? I126.CG2 F8.CD1 from 4.65 down to 3.38 Wed Jul 9 22:19:03 PDT 2008 Kevin Karplus try5-opt3 from modifying the initial model with the try1 loop, is doing much better than try4, which is optimized directly from TR432. rosetta, however, still likes try3-opt3.gromacs0.repack-nonPC better. Wed Jul 9 22:27:00 PDT 2008 Kevin Karplus I'm not sure what to do with this next. Ideally, I should come up with yet another loop conformation, perhaps from another server model. Servers that are close to TR432 (other than MUProt, which provided the starting point) include LEE-SERVER_TS1, Phyre_de_novo_TS1, and FEIG_TS1. Perhaps I should examine the loops on those, and make more chimeras. Wed Jul 9 22:39:03 PDT 2008 Kevin Karplus The FEIG_TS1 model looks a bit like our loops, but is missing 4 residues, which is cheating when modeling loops! LEE-SERVER_TS1 and Phyre_de_novo_TS1 have essentially the same loop as TR432. Thu Jul 10 14:50:38 PDT 2008 Kevin Karplus Perhaps the MULTICOM-REFINE_TS2 loop can be used as another starting point, copying A82-L95 from that into try4. Thu Jul 10 14:59:38 PDT 2008 Kevin Karplus On redoing the superposition, it is less different a loop than I thought, but I'll try it anyway. Thu Jul 10 15:06:16 PDT 2008 Kevin Karplus try6 will attempt to optimize chimera-try4-MULTICOM Thu Jul 10 16:25:14 PDT 2008 Kevin Karplus try6-opt3 doesn't score great, though much better than try4, which it was built from, and try6-opt3.gromacs0.repack-nonPC has the best rosetta energy so far. I've decided that I don't like try5 as much as I did initially, since I think it blocks access to conserved polar residues Y43,N81,Y85,N86, which are likely the active site. Maybe I should copy the more tightly packed C-terminus of try5 into try6, and optimize it again. Say copying I111-R130 from try5. Thu Jul 10 16:46:04 PDT 2008 Kevin Karplus try7 started to optimize chimera-try6-try5. Thu Jul 10 18:03:17 PDT 2008 Kevin Karplus OK, try7-opt3 now is the best scoring with the try7 costfcn, and rosetta likes try7-opt3.gromacs0.repack-nonPC best (by quite a bit). costfcn top 5 distinct models try1 try1-opt3 try5-opt3 try3-opt3 try7-opt3 try2-opt3 try2 try5-opt3 try3-opt3 try7-opt3 try2-opt3 try6-opt3 try3 try7-opt3 try3-opt3 try5-opt3 try6-opt3 try2-opt3.gromacs0.repack-nonPC try4 try7-opt3 try3-opt3 try5-opt3 try6-opt3 try2-opt3.gromacs0.repack-nonPC try5 try5-opt3 try7-opt3 try3-opt3 try6-opt3 try2-opt3 try6 try5-opt3 try7-opt3 try3-opt3 try6-opt3 try2-opt3 try7 try7-opt3 try5-opt3 try3-opt3 try6-opt3.gromacs0 try2-opt3.gromacs0 I think I've reached a point of diminishing returns---I'm unlikely to make things much better by fussing with the models any more. I'll submit: ReadConformPDB TR432.try7-opt3.pdb # < chimera-try6-try5 # best undertaker score (with try3,try4,try7.costfcn) ReadConformPDB TR432.try7-opt3.gromacs0.repack-nonPC.pdb # < chimera-try6-try5 # best rosetta energy ReadConformPDB TR432.try6-opt3.pdb # < chimera-try4-MULTICOM ReadConformPDB TR432.try5-opt3.pdb # < try3-opt3 < try2-opt3.gromacs0.repack-nonPC < chimera-init-try1 ReadConformPDB TR432.try4-opt3.repack-nonPC.pdb # < TR432 # optimized directly from provided model Thu Jul 10 18:42:53 PDT 2008 Kevin Karplus Submitted with comment: For this REFINEMENT model, we did a standard prediction for the residues included in the model, then tried optimizing both from the predicted model and from the provided starting point. Since we were directed to focus on Y85-G92, I also tried cutting and pasting that region from other models (particularly server models). This provided slightly different starting points and a bit more flexibility in rebuilding that loop. My tools are not optimized for fine-grain placement of atoms, so I'm not sure that I can make any improvement over a 1.34 Angstrom CA_RMSD model. All the models submitted have a disulfide constraint: C79-C101. The CYS residues seemed too close not to be interacting, though with this being part of a cytosolic human protein a disulfide somehow seems wrong. Since the cys are on the edge of the model, so I wonder if a metal-binding site in the protein has somehow been broken by taking out just this fragment. A disulfide here could correct for the missing metal. Model 1 TR432.try7-opt3.pdb # < chimera-try6-try5 # best undertaker score (with try3,try4,try7.costfcn) chimera-try6-try5: mostly from TR432.try6-opt3.gromacs0.repack-nonPC.pdb I111-R130 is from try5-opt3 chimera-try6-try5 was an attempt to use the tighter packing of the C-terminal helix in try5 with the more standard loop in try6. 2 TR432.try7-opt3.gromacs0.repack-nonPC.pdb # < chimera-try6-try5 # best rosetta energy Rosetta and undertaker disagree somewhat on clashes, so running an undertaker-produced model through gromacs energy minimization to relieve tiny clashes, then repacking the sidechains (except PRO and CYS) with Rosetta produces the best Rosetta energy scores, even though few atoms move much. 3 TR432.try6-opt3.pdb # < chimera-try4-MULTICOM chimera-try4-MULTICOM: mostly TR432.try4-opt3.repack-nonPC.pdb A82-L95 from MULTICOM-REFINE_TS2 This chimera was an attempt to get a slightly different loop as a starting point for further optimization. Sometimes doing this sort of cut-and-paste introduces breaks into the backbone and has slightly different shape to the fragments, kicking the optimization out of a local optimum and allowing it to find a different one. 4 TR432.try5-opt3.pdb # < try3-opt3 < try2-opt3.gromacs0.repack-nonPC < chimera-init-try1 chimera-init-try1: mostly TR432.pdb Y85-R97 from try1-opt3 try1-opt3 was the automatically generated prediction for TR432 It had a rather different idea of what the interesting loop should be. Initially I favored this model, but it looks like the loop now blocks access to the putative active site: Y43,N81,Y85,N86, so I downgraded the model. The try5 run had some extra constraints to try to pull the C-terminal helix in tighter, which were fairly successful, so I copied that portion of the model into chimera-try6-try5 to make the final model. 5 TR432.try4-opt3.repack-nonPC.pdb # < TR432 # optimized directly from provided model, using just undertaker. # helix constraints taken from the initial model and from try1-opt3 # disulfide C79-C101 Mon Nov 10 10:48:50 PST 2008 Kevin Karplus I did not improve GDT by refinement, nor real_cost.