Thu Jun 26 10:23:42 PDT 2008 T0474 Make started Thu Jun 26 10:24:11 PDT 2008 Running on cheep.cse.ucsc.edu Thu Jun 26 13:04:27 PDT 2008 Kevin Karplus The N-terminus (up to about L47) can be taken from a template (2ca9A, E-value 0.017), but the C-terminus seems to be more problematic. I'll do a polishing run on try1 to try squeeze it a bit tighter, but I think I'll then want to do more runs from alignments to get some different models. I'd like to bury W80, for example. Thu Jun 26 13:13:54 PDT 2008 Kevin Karplus try2 started to polish try1 try3 started to get another model from alignments Thu Jun 26 17:33:39 PDT 2008 Kevin Karplus With both the try2 and the try3 costfcn try2-opt3 beats try1-opt3 beats try3-opt3. Mon Jun 30 09:43:40 PDT 2008 Kevin Karplus Started MQAU1 and MQAC1 metaserver runs with try1 costfcn. pro-sp3-TASSER_TS1 is on top with both metaserver MQA functions, bus SAM-T08-server also does well as do the METATASSER and Zhang-Server models. If the SAM-T08-server model gets selected for both MQAU and MQAC, then I'll do an MQAX run that uses all servers EXCEPT SAM-T08. Mon Jun 30 11:05:11 PDT 2008 Kevin Karplus MQAU1 and MQAC1 do indeed choose SAM-T08-server_TS1, and don't score all that well in any case, so I'm doing an MQAX1 run also. Mon Jun 30 11:31:32 PDT 2008 Kevin Karplus MQAX1 seems to be working on SAM-T06-server_TS1, so I'll start an MQAY1 run also, that cuts out all the SAM servers. Mon Jun 30 14:30:57 PDT 2008 Kevin Karplus MQAY1 optimizes RBO-Proteus_TS3 Tue Jul 1 15:58:01 PDT 2008 SAM-T08-MQAO hand QA T0474 Submitted Tue Jul 1 15:58:01 PDT 2008 SAM-T08-MQAU hand QA T0474 Submitted Tue Jul 1 15:58:01 PDT 2008 SAM-T08-MQAC hand QA T0474 Submitted Sat Jul 5 09:57:01 PDT 2008 Kevin Karplus MQAX1 scores almost as well as try3, and MQAY1-opt3.gromacs0.repack-nonPC scores ok with the try3 costfcn also, though not as well as try1-opt3.gromacs0.repack-nonPC. Sat Jul 5 11:13:29 PDT 2008 Kevin Karplus Although MQAX1 scores well, I don't like the way it exposes so many predicted-buried residues. I should up the pred_near_back weight. try3 also seems a bit too open. Sat Jul 5 14:17:37 PDT 2008 Kevin Karplus try4 improves on try2, but I still don't like it that much. Let me try polishing try3 next. Sat Jul 5 14:22:28 PDT 2008 Kevin Karplus try5 started on peep. Sat Jul 5 15:04:38 PDT 2008 Kevin Karplus try5 improves on try3, but only just pulling ahead of MQAY1---still not up to the scores of try4 and MQAX1. The try4 series seems to be trying to form a small antiparallel sheet: I12-D15 to G64-I61, hbond L13 The try5 series seems to be trying to form a small parallel sheet: I12-D15 to G60-Q63 Hbond L13. I should try doing optimizations with each of these constraints added. Sat Jul 5 15:27:32 PDT 2008 Kevin Karplus try6 will attempt to improve try4 with the anti-parallel constraint. try7 will attempt to improve try5 with the parallel constraint. Sat Jul 5 17:06:45 PDT 2008 Kevin Karplus Rosetta likes best MQAX1-opt3.gromacs0.repack-nonPC try6 likes best try6-opt3, try4-opt3, try2-opt3, try7-opt3 try7 likes best try6-opt3, try4-opt3, try7-opt3 try2-opt3 Sat Jul 12 14:00:00 PDT 2008 Kevin Karplus Looking at one of the templates (2cpgA), I see that the sheet is formed across the dimer (chain A, chainB), so I need to model the dimer. I should probably make dimer models from each of the major tries, though I suspect that all of them have collapsed the monomers. I'll want a dimer constraint for M9-S17 antiparallel to S17-M9 (maybe G10 instead of M9): SheetConstraint M9 S17 S97 M89 Hbond S17 SheetConstraint G10 S17 S97 G90 Hbond S17 Sat Jul 12 14:29:18 PDT 2008 Kevin Karplus None of the dimers constructed from the monomers are any good--they all have horrendous clashes. I'll see if the dimer/try8 optimization can fix the problem. If not, I'll start over trying to make a monomer that does what I want. Sat Jul 12 15:54:07 PDT 2008 Kevin Karplus OOPS! I forgot to put "multimer 2" in all the try8 OptConform runs. Let me try dimerizing the 2 chains from try8-opt3 and do try9 with the cyclic multimer known in OptConform. Sat Jul 12 16:12:02 PDT 2008 Kevin Karplus dimer/try10 uses the same costfcn as dimer/try9, but attempts to start from alignments. I suspect that this will not work well, as it doesn't have the right relationship between the monomers to start from, so won't be able to pick out a good pair. dimer/try11 is similar to dimer/try10, except that it starts from dimer/try8-opt3 for the base of the TryAllAligns. Sat Jul 12 16:20:47 PDT 2008 Kevin Karplus The try11 run does not seem to be actually succeeding at trying anything. The problem is most likely that the initial model isn't really a cyclic dimer, and has few clashes, so none of the real cyclic dimers score as well (it is also likely that TryAllAlign is not making any attempt to create cyclic dimers) Sat Jul 12 18:48:53 PDT 2008 Kevin Karplus Sure enough, try11 never really got anywhere. try9-opt3 did the best, and try10-opt3.repack-nonPC scores ok too. I think that I may have the dimer constraint wrong! It looks to me like I want to pair GRILLDLS SLDLLIRG putting the charges on the same side of the sheet. SheetConstraint G10 L16 L96 G90 Hbond L16 For try12, I'll repeat try9, but with the new sheet constraint. For try13, I'll repeat try11, but with try9-opt3 as the initial model, and with the new sheet constraints. Sat Jul 12 21:20:27 PDT 2008 Kevin Karplus try12 scores better than try13, actually coming close to making the dimer. try12, try13, and try9 are essentially similar up to about Q50, then they vary enormously (well, try13 is essentially the same as try9, since it was modified from it). I need to look at longer members of this family, to see what the C-terminal helices are likely to do. Unfortunately, that seems to be about the domain boundary for a.43.1.3, so the adjacent domain may be anything (that dimerizes compatibly). Going down the best-scores list: The mainly beta domain in 2ca9A and 2bj7A (d.58.18.4) doesn't seem likely. 2cpgA, 2gpeA: short, nothing outside domain 1p94A: unsolved structure before domain, but nothing after 2bsqE: helix sticks out to interact with a different protein 2q2kA: short 2an7A: NMR models, floppy after first 2 helices. Given the floppiness of the 2an7A models, I'm not inclined to spend a lot of time on the region past N49. I could submit try12 and try13 to VAST, to see whether there is anything like them. Perhaps first I should make a try12-try13 chimera, with try12 for the N-termini and try13 for the C-termini (crossing over between I48 and N49), and polish it up to make a better example of the try9/try13 series. Sat Jul 12 21:58:40 PDT 2008 Kevin Karplus The chimera-try12-try13 scores well even before optimizing. I'll polish it (starting from all dimer models) in try14, and polish try12 (starting from all dimer models except chimera-try12-try13) in try15. Sat Jul 12 22:03:47 PDT 2008 Kevin Karplus I've started the VAST calculation with the chimera-try12-try13 model. Your VAST Search job was submitted at 07/13/2008 01:03:32(EDT). Request ID: 14904311062481363 VAST did not find any structure neighbors. Ah---but that was for a single chain. Perhaps if I named the chain right as chain A and chain B, it would work better. Sat Jul 12 22:07:21 PDT 2008 Kevin Karplus Your VAST Search job was submitted at 07/13/2008 01:06:49(EDT). Request ID: 1044907666137404363 Sat Jul 12 22:08:15 PDT 2008 Kevin Karplus I just noticed that try14 was polishing try12, not the chimera, so I'll start a try16 with the same costfcn, but excluding try12 models. Sat Jul 12 22:11:40 PDT 2008 Kevin Karplus Even with the separate chains, VAST couldn't find similar structures---I guess one strand and 2 helices aren't distinctive enough for it. Sun Jul 13 06:56:10 PDT 2008 Kevin Karplus try16 now scores best, followed by try14, then try15. try14 did eventually swtich over to using chimera-try12-try13, but try16 had more polishing. OK, I'm ready to submit ReadConformPDB T0474.try16-opt3.unpack.gromacs0.repack-nonPC.pdb chain A # < chimera-try12-try13 ReadConformPDB T0474.try14-opt3.unpack.pdb chain A # < chimera-try12-try13 ReadConformPDB T0474.try15-opt3.unpack.pdb chain A # < try12-opt3 ReadConformPDB T0474.try12-opt3.unpack.pdb chain A # < dimer-try3-2cpg ReadConformPDB T0474.try13-opt3.unpack.pdb chain A # < try9-opt3 I only have two really distinct models here (the try12 or try13 C-termini). I probably should have more, but I suspect that the molecule is floppy after N49. Sun Jul 13 07:31:53 PDT 2008 Kevin Karplus I made find-try16.under to look through the server models for something close to try16. The closest is BAKER-ROBETTA_TS1, and it isn't clear that it would dimerize. If I had a lot of time, I'd probably make dimers of several of the metaserver models and optimize them.