Thu Jun 1 09:05:08 PDT 2006 T0310 Make started Thu Jun 1 09:28:57 PDT 2006 Running on lopez.cse.ucsc.edu Thu Jun 1 09:30:14 PDT 2006 Kevin Karplus Superb blast hit to 1vluA (39% id over 452 residues). Also good hit to 1o20A (34% id over 410 residues). Thu Jun 1 21:44:48 PDT 2006 Kevin Karplus There are many hits to c.82.1.1 and c.82.1.2 (29 in best-scores.rdb). 1o20A scores on top, and 1vluA is way down on the list, probably because it isn't in the template libraries All the top hits in T0310.undertaker-align look essentially the same. I think we may need to make a dimer to optimize the fine details. 1vlu[AB] may be an appropriate dimer to use (but use the biological unit, not the asymmetric unit). We may want to look through the 29 hits to find a PDB file that has the dimer organized the way we want it. Thu Jun 1 22:24:21 PDT 2006 Kevin Karplus We have comparative modeling up to about H400, but after that---who knows what the last 25 residues do? It might help to have the dimer---at least then we can keep the extra residues from running through the other monomer. Sat Jun 17 14:23:13 PDT 2006 Kevin Karplus It looks like 1euhA may be a good template for building the TETRAMER. I made make-dimer1.under and created a tetramer use try1-opt2. It is clear that the C-termini pack together in the center of the 4mer, so we have to move the extra C-terminal stuff out of the way. Currently T425 is buried deep in a neighboring monomer. I think that we may be able to improve things by forcing L414.CA-L94.CA to be around 30 (or more), instead of around 6; W393.CA-L411.CA to be around 30 (or more), instead of around 13; T425.CA-D399.CA to be around 40 (or more), instead of around 33. I'll try this as try2, and then make a new tetramer. Sat Jun 17 14:54:42 PDT 2006 Kevin Karplus try2 started on whidbey. Sat Jun 17 16:17:59 PDT 2006 Kevin Karplus Somewhat surprisingly, ReduceConstraint was never successful in making try2-opt1. I wonder if there is a bug in it, since there certainly should have been fragment insertions capable of reducing the hand-added distance constraints. Sat Jun 17 19:23:58 PDT 2006 Kevin Karplus try2 was a complete waste of time---the C-terminal end was not moved out where I wanted it---possible because that would look like a huge increase in the radius of gyraton. Perhaps I can ask Firas to ProteinShop the part after D399, so that it sticks straight out from the end of the sheet. Sat Jun 17 21:49:03 PDT 200 Firas Khatib I did this and it is in: /casp7/T0310/decoys/Proteinshop c-termOut.renum.pdb - starts where the coil is at 397 & c-termOut399on.renum.pdb - starts at 399, by making 397 & 398 a strand extended even more is this one: c-termOut399onExtend425.renum.pdb Sun Jun 18 06:48:22 PDT 2006 Kevin Karplus Thanks, Firas. script will use to make -------- -------------- -------------- make-tetramer3.under c-termOut.renum.pdb tetramer3-1euhA.pdb make-tetramer4.under c-termOut399on.renum.pdb tetramer4-1euhA.pdb make-tetramer5.under c-termOut399onExtend425.renum.pdb tetramer5-1euhA.pdb Sun Jun 18 07:22:54 PDT 2006 Kevin Karplus All of the extensions look fine (the distance constraints were not really that important---I was just trying to get something vaguely like any of the ones Firas made). Of the tetramers, tetramer4-1euhA scores best, but they are all quite close in score, and all clearly have the C-terminus wrong, just not in the way any more. I will try doing a short optimization run from the tetramers adding a sheet constraint to hold T390-K392 of chain A parallel to H378-R380 of chain B. (tetramer/try1 started on whidbey) Sun Jun 18 07:56:31 PDT 2006 Kevin Karplus The tetramer try1 optimization is failing badly, probably because the monomers were not put in order so that A->B->C->D->A were all the same transformation. (We have A<->B and C<->D). I'll rearrange the T0310-1euhA.tetramer-a2m file, remake the tetramers, and rerun tetramer/try1 I'll have to check that I get the sheet constraint between monomers properly fixed. Sun Jun 18 08:20:57 PDT 2006 Kevin Karplus Bad idea. The symmetry group for 1euh seems to be S_{2,2} not S_4, so I'm better off statying with the original numbering, and pretending that it is a dimer. Sun Jun 18 09:02:56 PDT 2006 Kevin Karplus Treating the tetramer as a pair of dimers does seem to work ok, with OptSubtree and TweakMultimer serving as the two operators that adjust the relative positioning of the monomers. Sun Jun 18 09:59:47 PDT 2006 Kevin Karplus Optimizing the tetramer is *very* slow. I may need to start another optimization run with fewer generations and fewer iterations, just to get something to complete (even though I already reduced the size considerably from the standard). Sun Jun 18 17:20:35 PDT 2006 Kevin Karplus I think that tetramer/try1 has moved the pairs of dimers apart, but try1-opt1 looks at least plausible. Sun Jun 18 20:53:41 PDT 2006 Kevin Karplus I made a monomer from chain A of tetramer/ try1-opt1 (since it will take a long time to finish making try1-opt2) and am optimizing it separately as try3. I think that the distance constraints should be enough to keep the linker region from crumpling up into where a neighboring monomer would be, but I may be wrong. Mon Jun 19 16:43:20 PDT 2006 Kevin Karplus try3-opt2.repack-nonPC scores better than try3-opt2, but both score better than previous monomers. The extended part has been wound into a helix, which seems ok to me. make-tetramer6.under takes try3-opt2.repack-nonPC and makes a tetramer out of it. Unfortuantely, the helix collides badly with the neighboring monomer. Perhaps I can get Firas to swing the helix around 180 degrees, so that it would just contact the monomer on the other side. Mon Jun 19 17:28:23 PDT 2006 Kevin Karplus Firas made two attempts at the rotation: decoys/Proteinshop/rotate180.renum.pdb decoys/Proteinshop/rotateTake2.renum.pdb I made these into tetramers tetramer/decoys/tetramer7-1euhA.pdb tetramer/decoys/tetramer8-1euhA.pdb The second one tetramer/decoys/tetramer8-1euhA.pdb looks feasible. Mon Jun 19 17:47:38 PDT 2006 Kevin Karplus I started tetramer try2 on cheep. It should be a fairly short run (less than 5 hours). Tue Jun 20 12:40:53 PDT 2006 Kevin Karplus Well, it was over 8.5 hours on cheep, but the results are not bad. tetramer/try1-opt2 scores a bit better, but I prefer the final helix on try2-opt2. I'll submit those 2 predictions (try2-opt2 first) as "dimers", and choose as monomers chain A of tetramer try2-opt2,unpack chain A of tetramer try1-opt2.unpack try3-opt2.repack-nonPC try2-opt2.repack-nonPC align 1 of undertaker-align (from 1o20A) Tue Jun 20 13:04:00 PDT 2006 Kevin Karplus Preliminary submission of monomers done. Tue Jun 20 13:04:35 PDT 2006 Kevin Karplus I'll have to make a "dimer" directory to make the "half-tetramer" submissions. Tue Jun 20 13:13:56 PDT 2006 Kevin Karplus chains A+B of tetramer try2-opt2.unpack chains A+B of tetramer try1-opt2.unpack submitted as a "dimer" Tue Jun 20 14:47:17 PDT 2006 Kevin Karplus Some of the submissions were rejected, because the labeling was wrong---the problem was that bout dimer and tetramer Makefiles had the MOMOMER_LENGTH wrong. Tue Jun 20 15:14:52 PDT 2006 Kevin Karplus Luckily, the MONOMER_LENGTH affected only the unpacking scripts, not the optimization, so I was able to unpack again and get the submission accepted. Since we now have 5-6 days, it might be worth doing some more optimization of tetramer try2-opt2 to reduce clashes further. Tue Jun 20 16:20:20 PDT 2006 Kevin Karplus started tetramer/ try3 on the farm cluster. Sat Jun 24 10:34:46 PDT 2006 Kevin Karplus tetramer/ try3 did not make much change. We don't have time for more work and the model is decent, so I'll just submit the tetramer try3/try2/try1 multimers and monomers, with the try3 and try2 monomers as models 4 and 5. Fri Jul 14 12:00:23 PDT 2006 Kevin Karplus making of evaluate.unconstrained.pretty still running, after 14.6 hours. Thu Mar 22 08:43:16 PDT 2007 Kevin Karplus Evaluation now takes only 5567 seconds. This target was cancelled by the assessors, but I'm not sure why. Our best model was try1-opt1, but there were several servers that did better, including SAM-T99_AL4 and SAM-T02_AL1 (when scwrled). The best we submitted was model5=try2-opt2.repack-nonPC, which was done purely as a monomer (the c-terminus conflicts badly as a tetramer). The 2h5g structure appears to be 2 dimers, rather than a tetramer. It isn't a dimer in the unit cell---the dimer is formed by a symmetry operation. I'm not sure why we tried to build a tetramer, as the "additional information" in T0310.doc.html clearly says that there is a dimer.