Fri May 16 09:41:09 PDT 2008 T0405 Make started Fri May 16 09:42:01 PDT 2008 Running on cheep.cse.ucsc.edu Fri May 16 09:57:13 PDT 2008 Kevin Karplus BLAST does not find anything for this 282-residue target. The crystallographers report "1x sulphate bound, pH 6.5, monomeric, no disulfide bridges". Fri May 16 13:37:01 PDT 2008 Kevin Karplus ORFan! this one will be a bit hard to predict. The top hit is 2ii1A at E-value 6. We're going to be reduced to new-fold methods. Fri May 16 17:16:29 PDT 2008 Kevin Karplus Our first model try1-opt3 does not look too bad. The strand P83-N88 could be an extra strand, maybe StrandConstraint P83 N88 V157 R153 hbond V154 Fri May 16 21:54:08 PDT 2008 Kevin Karplus Oops. I goofed. I ran try2, but I meant to say SheetConstraint P83 N88 F158 R153 hbond V154 Let me try again for try3. Sat May 17 08:20:08 PDT 2008 Kevin Karplus try3 did not succeed at adding the strand---try1 still comes closest. try3 does not score very well, getting neither the try1 nor the try2 sheets. The try3 sheet pairing looks like an unlikely one, as the hairpin has a huge gap between R167 and V168. Sat May 17 15:07:59 PDT 2008 Kevin Karplus The try4 costfcn still likes try2-opt3.gromacs0.repack-nonPC best, which implies to me that it may be paying too much attention to breaks and clashes, or that the constraints are too inconsistent. Mon May 26 14:22:39 PDT 2008 Kevin Karplus The MQAU costfcn likes the Zhang-Server and BAKER-ROBETTA best. This one is going to be tough. Mon May 26 21:22:29 PDT 2008 Kevin Karplus I like the Zhang-Server_TS1 bundle for D198 to the end. I like BAKER-ROBETTA_TS1 for Y82-N183 I need to look at servers/Zhang-Server_TS5 servers/Zhang-Server_TS2 servers/BAKER-ROBETTA_TS3 Wed May 28 11:13:04 PDT 2008 SAM-T08-MQAO hand QA T0405 Submitted Wed May 28 11:13:04 PDT 2008 SAM-T08-MQAU hand QA T0405 Submitted Wed May 28 11:13:04 PDT 2008 SAM-T08-MQAC hand QA T0405 Submitted Mon Jun 9 09:30:28 PDT 2008 Kevin Karplus I think I should run the metaserver optimizations with just the try1 costfcn, since I really have no idea what to expect here. The MQA functions are predicting GDT around 22%-31%, so the metaservers are not very convincing either. Mon Jun 9 19:39:26 PDT 2008 Kevin Karplus MQAC1-opt3 is based mainly on SAM-T08-server_TS1 Tue Jun 10 10:41:20 PDT 2008 Kevin Karplus MQAU1-opt3 is based mainly on BAKER-ROBETTA_TS3 The try4 costfcn prefers MQAC1-opt3.gromacs0 MQAU1-opt3.gromacs0 try2-opt3.gromacs0 try1-opt3.gromacs0 try4-opt3.gromacs0 try3-opt3 The try1 costfcn prefers MQAC1-opt3 MQAU1-opt3 try1-opt3,repack-nonPC try2-opt3 try3-opt3 try4-opt3.gromacs0.repack-nonPC Fri Jun 13 08:24:28 PDT 2008 Kevin Karplus The try5-opt3 model is not very convincing, but it forms a lot of sheet. It scores worse than try1-try4 and the MQAC1 MQAU1 models, even with the try5 costfcn. Sun Jun 15 09:52:10 PDT 2008 Kevin Karplus Zhang-Server_TS5 seems to be 3 repeated 4-helix bundles, with extended conformation but no hbonds for F152-L181, which is predicted to be a meander. I don't like it at all. Zhang-Server_TS2 sort of forms the meander (one strand is not Hbonded in place) and has nice conformations for the helix bundles, getting good burial for the hydrophobic faces. BAKER-ROBETTA_TS3 gets the 3-strand meander, and almost adds another strand to the sheet. The helices are not as neatly packed as the Zhang-Server models. Maybe what I need to add to undertaker is the ability to build a fragment library from a collection of conformations (and not just largest gapless segments, but any length fragment), so that I can more easily mix-and-match super-secondary structure elements. The current CrossOver operator is too limited. Even a "DoubleCross" operator, that took two break points and made an A-B-A model would be more powerful than what I currently have. Sun Jun 15 10:31:56 PDT 2008 Kevin Karplus OK, I made a quick stab at a DoubleCross operator and will try it out for try6. Oops, I mis-spelled it in the InitMethodProbs list, so that DoubleCross gota pseudocount of 0.1 instead of 50, and was rarely tried, instead of frequently tried. Oh, well, I'll have to try again in try7. Sun Jun 15 20:18:45 PDT 2008 Kevin Karplus try6-opt3, based mainly on MQAC1-opt3 from SAM-T08-server_TS1, still has breaks and clashes, particularly breaks before K228, G192, R167, and M227. Wed Jun 18 08:16:29 PDT 2008 Kevin Karplus try7 started on moai cluster to polish try6. Should I create a new costfcn that tries to attach L84-V87 antiparallel to F152-F158? If so, with what phase? Let's try SheetConstraint L84 V87 L156 R153 hbond L156 I stuck this in with the try6 sheets and helices to try optimizing from alignments for try8. Wed Jun 18 09:14:37 PDT 2008 Kevin Karplus Big question: should I try grabbing chunks from the metaserver models and try assembling them differently? Wed Jun 18 17:34:41 PDT 2008 Kevin Karplus Let me try making a chimera, with P81-T186 from BAKER-ROBETTA_TS1 D194-Q282 from Zhang-Server_TS1 M1-C75 from Zhang-Server_TS2 linker regions D76-Y80 and E187-Y193 from MQAC1-opt3.gromacs0 I'll try optimizing this chimera-baker-zhang with the sheet constraints from the BAKER-ROBETTA_TS1 model and the helix constraints from the appropriate parts of the two Zhang-Server models. I'll also add SheetConstraint K111 P115 R153 V157 Hbond V154 to try to close up the sheet that seems to be forming. Wed Jun 18 18:58:00 PDT 2008 Kevin Karplus try9-opt1 seems to be having some trouble getting the pieces to fit together. The C-terminal domain looks like it needs to rotate 90 degrees to fit in place, and the N-terminal domain looks like it got pulled apart a bit. I may need to make another chimera, starting with this one and overlaying the helix bundles again. I might also want to add some bundling constraints to hold the bundles together. Wed Jun 18 20:44:46 PDT 2008 Kevin Karplus Before trying to make chimeras or tweak try9, I'll do a polishing run to try to reduce the breaks. Thu Jun 19 08:19:52 PDT 2008 Kevin Karplus try10-opt3 is now getting up close to MQAU1 and try7. strand L84-V87 has gotten detached from the sheet, and the N-terminal domain messed up a bit, and I think that the C-terminal domain needs rotating, to pack L214 near K129, Q211 near F133, L207 near W136, and E203 near E140. I'll make a new chimera chimera-try10-baker1 that takes try10-opt3.gromacs0 except for P83-L118 from BAKER-ROBETTA_TS1. I'll make another chimera with the N-domain from Zhang-Server_TS1, but it is going to have clashes. To keep the domain together, I'll add helix constraints for K57-K71, E28-K51, and M1-T22, and distance constraints M1.CA T58.CA 5.7 T4.CA T58.CA 6.0 V8.CA V44.CA 5.4 I45.CA T58.CA 6.8 F65.CA A9.CA 7.2 L66.CA I41.CA 7.2 L69.CA F16.CA 9.6 L69.CA I38.CA 9.7 The chimera-try10-baker1-zhang2-zhang1 model has the C-terminal domain rotated, but it is now 180 degrees out of position, rather than 90 degrees. I hope that it can be flipped by undertaker! Thu Jun 19 09:49:31 PDT 2008 Kevin Karplus For try11, I added packing constraints to keep the N and C terminal domains together, and added some placement constraints to try to get the C-terminal domain oriented the way I believe it goes. I also simplified the helix constraints, since I don't think that the helices necessarily have kinks. Thu Jun 19 10:58:24 PDT 2008 Kevin Karplus try11 crashed after try11-opt output, with undertaker: Spot.cc:172: XYZpoint Spot::get_spot(const ChainsResiduesAndAtoms*, int, const Pointlist&, const Conformation*, bool, const Spot*) const: Assertion `return_point.mag2() < 1.e12' failed. (probably the same bug I've been trying to track down that very occasionally returns bad Transform from the constructor. In any event, try11-opt2 did not hold the Nterminal domain together and did not rotate the C-terminal domain the way I wanted. I'll up the corresponding weights for try12 and try again. Thu Jun 19 12:00:37 PDT 2008 Kevin Karplus try12-opt3 is still not placing the C-terminal domain where I want. I'll try a different set of constraints for C.place and upping the weights for try13. Thu Jun 19 15:33:12 PDT 2008 Kevin Karplus try13 did not rotate the C-terminus as I had hoped, but just pulled one of the helices apart a bit. try13 does pack the N-terminal helices in closer. Thu Jun 19 16:02:16 PDT 2008 Kevin Karplus For try14 I'll just try polishing the models, since try12 and try13 score best with the try14 costfcn. Thu Jun 19 16:29:01 PDT 2008 Kevin Karplus I'll also do a try15, with the try13 costfcn, but starting only from chimera-try10-baker1-zhang2-zhang1 which is further from optimized, so should be easier to move around. Thu Jun 19 21:19:24 PDT 2008 Kevin Karplus try15 looks pretty messed up, but the C-terminal helices may be in about the right spot. Perhaps I can superimpose some pieces and make a new chimera that does the right thing. First I'll make three chunks: P81-T186 from try14 BAKER-ROBETTA_TS1 D194-Q282 from Zhang-Server_TS1 M1-C75 from Zhang-Server_TS2 Thu Jun 19 21:37:52 PDT 2008 Kevin Karplus I tried superimposing those chunks on try15, and realized that there is a huge, uncloseable gap between D194 and V195 if we put the C-terminal 4-helix bundle in aligned with the helices of try15. I'd need a very different conformation of T186-D194 to close the gap. I could try tweaking up M1-N72 from zhang2 Q73-Y82 from try15 P83-T186 from baker1 E187-Y193 from try15 D194-Q282 from zhang1 I put that together as chimera-try15, but I don't think it is worth the trouble of trying to fix. I'll now try to get 5 very different models to submit: 1 try14-opt3.gromacs0 < try12-opt3.repack-nonPC < chimera-try10-baker1 try10 < try9-opt3.repack-nonPC < from Zhang-Server_TS2, BAKER-ROBETTA_TS1, Zhang-Server_TS1 2 try7-opt3 < try6-opt3 < MQAC1-opt3 < SAM-T08-server_TS1 3 MQAU1-opt3.gromacs0 < BAKER-ROBETTA_TS3 4 try2-opt3.gromacs0.repack-nonPC < 2f4eA 5 try4-opt3 < 2aaa