Wed Jun 21 10:32:14 PDT 2006 T0338 Make started Wed Jun 21 10:34:53 PDT 2006 Running on shaw.cse.ucsc.edu Wed Jun 21 10:43:18 PDT 2006 Kevin Karplus Blast gets encouraging, but not definitive full-length hits in PDB: 1kxu 23% id over 211 residues 2.2e-04 1jkw 23% id over 211 residues 2.2e-04 There are other, somewhat weaker hits. The HMMs may help get better alignments. Wed Jun 21 19:24:59 PDT 2006 Kevin Karplus The HMMs get strong hits to paired a.74.1.1 domains and a.47.1.2 domains (the top blast hits are a.74.1.1 domains). The best-scoring hit with the HMMS is 2cchB. Mon Jul 10 11:39:55 PDT 2006 Pinal Kanabar I increased wt for break and soft clashes. Added a new HelixConstraint from A101 to L108. I am hoping that the strange turn will look better. Running try2 on camano Summary.html : We have top 2cchB 261 4.6955e-30 1w98B 271 1.0575e-23 1vin 268 7.0472e-20 1vin a.74.1.1 a.74.1.1 (all alpha protein) 1kxu 333 1.5177e-16 1jkw a.74.1.1 a.74.1.1 Top hit 2cchB has it seems four chains :it looks like it has two diffrent dimers. try1.pdb looks similar to one of the dimer and has all alpha helices. Mon Jul 10 16:29:37 PDT 2006 Pinal Kanabar I guess by increasing wt I made try2 worst... In try3.costfnc I will keep all the try1.costfcn wts commented : #include T0338.dssp-ehl2.constraints #include T0338.undertaker-align.sheets #include rr.constraints took try1-opt2.helices and removed the one which I didnt like -- Yet to finish try3.costfcn Tue Jul 11 11:11:16 PDT 2006 Firas Khatib we just ran try3 on vashon, it has these changes to the constraints: #include T0338.dssp-ehl2.constraints #include T0338.undertaker-align.sheets include rr.constraints #from try1-opt2.helices HelixConstraint (T0338)A1 (T0338)E10 1 HelixConstraint (T0338)T8 (T0338)E13 1 HelixConstraint (T0338)E27 (T0338)N46 1 HelixConstraint (T0338)S48 (T0338)H66 1 HelixConstraint (T0338)N72 (T0338)E88 1 #HelixConstraint (T0338)K93 (T0338)A101 1 #want to make this helix longer HelixConstraint (T0338)L94 (T0338)L105 3 #HelixConstraint (T0338)P107 (T0338)K115 1 #do not want previous helix #HelixConstraint (T0338)E125 (T0338)G138 1 #want to make this helix longer HelixConstraint (T0338)C116 (T0338)L137 3 HelixConstraint (T0338)H145 (T0338)V151 1 HelixConstraint (T0338)D149 (T0338)T154 1 HelixConstraint (T0338)K152 (T0338)R158 1 HelixConstraint (T0338)S160 (T0338)S174 1 HelixConstraint (T0338)K185 (T0338)S201 1 HelixConstraint (T0338)S208 (T0338)W215 1 HelixConstraint (T0338)T223 (T0338)Q236 1 HelixConstraint (T0338)N243 (T0338)N250 1 Tue Jul 11 13:25:34 PDT 2006 Pinal Kanabar I wanted to get model from my best pdb hits . try4.under : InfilePrefix 1kxu/ include read-alignments-scwrl.under InfilePrefix 1vin/ include read-alignments-scwrl.under in try4.costfcn : used alignment of try1.under Its running on orcas Tue Jul 11 13:36:40 PDT 2006 Firas Khatib the next run, when try3 ends, will have this constraint in it: Constraint I38.CB V97.CB -10. 7.0 14.0 0.85 Tue Jul 11 15:08:10 PDT 2006 Pinal Kanabar try4 scores the best but it has long helix which doesn't seem correct -- I dont like it. try3 is finished and as we decided earlier for try5 we increased 'constraints 20' and changed rr constraint in a hope that it will bring helix near to the core. It is running on orcas. try5.costfcn : #keep rr constraints, but not the bonus ones Constraint I76.CB M133.CB -10. 7.0 14.0 0.601545537313 Constraint H60.CB S77.CB -10. 7.0 14.0 0.574932381932 Constraint M59.CB L83.CB -10. 7.0 14.0 0.552654894108 Constraint L83.CB M133.CB -10. 7.0 14.0 0.550353958212 Constraint A55.CB S77.CB -10. 7.0 14.0 0.542377608627 Constraint E130.CB F139.CB -10. 7.0 14.0 0.531776376545 Constraint A80.CB M133.CB -10. 7.0 14.0 0.516473237024 #change this bonus to real constraint so keep helix closer to core Constraint I38.CB V97.CB -10. 7.0 14.0 1.5 Tue Jul 11 16:53:03 PDT 2006 Firas Khatib we made a dimer using 1kxu and try1. It doesn't work well due to residues 120-126 in try1 sticking out and messing up the dimer interface. We will try the same dimer using try3 instead Tue Jul 11 17:17:02 PDT 2006 Firas Khatib try3 works MUCH better as a dimer! try1 is running in the dimer directory as mult1 on lopez, it is using try3. as soon as try5 ends I will make a dimer from that and optimize it Tue Jul 11 17:41:56 PDT 2006 Kevin Karplus I need a list of models both for the monomer and the dimer before 8pm. I'm not awake enough to guess which models to submit based on the notes so far in this README file. Tue Jul 11 19:49:55 PDT 2006 Firas Khatib I will get you that in 10 minutes! for the monomer, the order in superimpose-best.under is still valid: T0338.try3-opt2.pdb T0338.try1-opt2.pdb T0338.try4-opt2.pdb T0338.try5-opt2.pdb T0338.try2-opt2.pdb for the dimer, sadly... as of 8pm, we only have try1 in the dimer/decoys directory for the soft deadline: T0338.try1-opt2.pdb.gz which comes from dimer-1kxu-try3-opt2.pdb do not submit dimer-1kxu-try1-opt2.pdb as it is garbage. Tue Jul 11 20:07:31 PDT 2006 Firas Khatib dang, try5 does not work as a dimer at all. Tue Jul 11 20:38:21 PDT 2006 Kevin Karplus Instead of try3, I'll use dimer/decoys/T0338.try1-opt2.unpack.gromacs0.repack-nonPC.pdb chain A which is try3 optimized in a dimer context. Instead of try5, I'll use try5-opt2.gromacs0.repack-nonPC.pdb which is Rosetta's favorite. Tue Jul 11 20:43:58 PDT 2006 Kevin Karplus Preliminary monomer submission done: InFilePrefix dimer/decoys/ ReadConformPDB T0338.try1-opt2.unpack.gromacs0.repack-nonPC.pdb chain A InFilePrefix decoys/ ReadConformPDB T0338.try1-opt2.pdb ReadConformPDB T0338.try4-opt2.pdb ReadConformPDB T0338.try5-opt2.gromacs0.repack-nonPC.pdb ReadConformPDB T0338.try2-opt2.pdb Tue Jul 11 21:06:23 PDT 2006 Kevin Karplus For dimers, I submitted dimer/ try1-opt2 dimer/ try1-opt2.unpack.gromacs0.repack-nonPC Wed Jul 12 14:32:22 PDT 2006 Firas Khatib try4 looks different enough at the N-terminus to try to dimerize. the dimer interface looks decent enough, pinal will optimize it as try3 in the dimer directory I am going to have another crack at try5. It had all the constraints from try3 except that I had turned off the rr bonuses and changed 1 from bonus to really strong. It did what I wanted with that 1 constraint but sadly put the N-terminus in the wrong place, like try2. I have started try6 to have the EXACT same constraints as try3 with that one bonus constraint changed to a strong one. try6 is running on vashon Wed Jul 12 16:54:23 PDT 2006 Firas Khatib try6-opt1 is different enough from try4 and try3 to try dimerizing. when opt2 finishes I will dimerize it and optimize that as try4 Wed Jul 12 21:44:53 PDT 2006 Firas Khatib dang, the dimer interface is where the try6 difference is and this will not work at all! Thu Jul 13 10:32:14 PDT 2006 Pinal Kanabar I optimized try3 and it doesnont seem to work. I looked at dimer/decoys/T0338.try3-opt2.gromacs0.pdb.gz chains are really far from each other. There is hardly any interface between them ... I dont what else should be tried... Thu Jul 13 13:17:10 PDT 2006 Firas Khatib yes indeed, it seems that the interface from dimer-1kxu-try3-opt2 or any of the other ones doesn't seem to work when optimized. I opened them in Proteinshop and there are a few clashes on the interface, but not that many. I will try to see why decoys/T0338.try1-opt2.pdb.gz is based on decoys/dimer-1kxu-try3-opt2.pdb and decoys/T0338.try3-opt2.pdb.gz is based on decoys/dimer-1kxu-try4-opt2.pdb look nothing like their starting points. in dimer: try4 will be a copy of try3, but with: InFilePrefix ../ ReadFragmentAlignment T0338.t2k.many.frag ReadFragmentAlignment T0338.t04.many.frag ReadFragmentAlignment T0338.t06.many.frag instead of InFilePrefix ReadFragmentAlignment T0338.t2k.many.frag ReadFragmentAlignment T0338.t04.many.frag ReadFragmentAlignment T0338.t06.many.frag hopefully that was one of the problems, otherwise I will add in constraints for the dimer interface. try4 is running on orcas Pinal is going to launch try5 which will be a copy of try1, but with the same fix try5 is running on vashon Thu Jul 13 14:59:35 PDT 2006 Firas Khatib before try4-opt2 is done try3 is still scoring the best in the dimer directory (although try4-opt1 looks promising) so I will run try3 again (which is what try4 is based on as well) but with soft_clashes turned down in case that is what is causing the dimer interface to fall apart. soft_clashes will be 10 instead of 20 recapping: try3,try4,try6 are all based off dimer-1kxu-try4-opt2 try1 and try5 are both based off dimer-1kxu-try3-opt2 try6 running on whidbey Thu Jul 13 15:56:22 PDT 2006 Pinal Kanabar Created dimer based on 1zp2.mmol : dimer-1zp2-try3-opt2.pdb . It looks good but it has little clash but we decided to go ahead and optimize it. Running try7 on camano. Thu Jul 13 16:20:39 PDT 2006 Firas Khatib I Proteinshopped the clash and hope that will help. this will be try8. recapping: try3,try4,try6 are all based off dimer-1kxu-try4-opt2 try1 and try5 are both based off dimer-1kxu-try3-opt2 try7 and try8 are both based off dimer-1zp2-try3-opt2 try2 never ran because the dimer looked horrible. so far, I like try5 the best Thu Jul 13 16:31:30 PDT 2006 Firas Khatib try8 is running on shaw Fri Jul 14 11:29:26 PDT 2006 Firas Khatib try8 is the only run we have where the dimer interface is actually in the same spot as the dimer interface from the parent. sadly, is scores much worse than all the other trys! we will optimize try8 using ReadConformPDB with try8 as input. this will be try9 since the score-all.try8.pretty has try8 with really bad phobic_fit, we increased phobic_fit from 2 to 4 in try9.costfcn that is the only difference and we will see if that helps or not. try9 is running on camano. recap: try1,try3,try4,try5,try6 all based off dimer-1kxu try7,try8 both based off dimer-1zp2-try3-opt2 try9 based off try8 try5 is the best dimer based off 1kxu try8 is the best dimer based off 1zp2 Fri Jul 14 12:00:48 PDT 2006 Firas Khatib we are going to do the same thing with try5 (as try9) try10 will be based off try5 with phobic_fit set at 4 instead of 2. try10 is running on lopez Fri Jul 14 15:28:26 PDT 2006 Firas Khatib try10 seems like the only interface that seems good it scores badly due to phobic_fit, but after discussion with Kevin we decided that turning phobic_fit off would be worth a try. This will be try11 running on camano try12 will be based on try8, but with phobic_fit turned off. recap: try12->try8->dimer-1zp2-try3-opt2.pdb try11->try5->dimer-1kxu-try3-opt2.pdb Fri Jul 14 17:41:02 PDT 2006 Firas Khatib try11 and try12 did not work because the line was left blank so everything after did not get used in the costfcn! try11 will be rerun as try14 and try12 will be try13 running on shaw & lopez I am going to take try10 and give it constraints on the dimer interface this will be try15 running on vashon. recap: try13->try8->dimer-1zp2-try3-opt2.pdb try12->try8->dimer-1zp2-try3-opt2.pdb try14->try5->dimer-1kxu-try3-opt2.pdb try11->try5->dimer-1kxu-try3-opt2.pdb try15->try10->try5->dimer-1kxu-try3-opt2.pdb Fri Jul 14 21:34:08 PDT 2006 Kevin Karplus I'm not sure which models I'm supposed to submit as monomers or as dimers. There are only 3 monomers in superimpose-best.under, and I need 5. There are only 2 dimers in dimer/superimpose-best.under, and I need 5. I'll stick a bunch of models into dimer/superimpose-best.under, and try to make sense of them, but I'd *really* like it if this stuff was presented to me so that I could work on new targets, rather than spending hours having to root around trying to figure out what other people have done! Fri Jul 14 21:56:20 PDT 2006 Firas Khatib I'll have the final superimpose-best.under files soon! Fri Jul 14 22:11:09 PDT 2006 Kevin Karplus The dimers from 1zp2 don't seem to have ever been optimized with inter-monomer constraints. It looks like we have really only 2 different dimers, those based on 1kxu (or 1jkw) and those based on 1zp2. Most of the 1zp2 ones have drifted away, so only dimer/try8 of them is useful. Of the 1kxu ones, the best is try15-opt2 (or try15-opt2.unpack.gromacs0.repack-nonPC, which is rosetta's favorite) other reasonable different choices include dimer-1kxu-try3-opt2 try4-opt2 dimer-1jkw-try4-opt2 Unless Firas comes up with some I missed, for the dimers I propose try15-opt2.unpack.gromacs0.repack-nonPC # rosetta favorite, 1kxu try3 monomer try8-opt2 1zp2 try4-opt2 1kxu try4 monomer dimer-1kxu-try3-opt2 dimer-1jkw-try4-opt2 Fri Jul 14 22:49:56 PDT 2006 Firas Khatib I am re-running try8 with the following inter-monomer constraints: Constraint T477.CB P220.CG -10. 4.0 6.0 5 Constraint Q492.CD Q236.CD -10. 4.0 6.0 5 this is try16 on shaw. I am also running the same constraints with a proteinshopped file as input: dimer/Proteinshop/dimer-1zp2-try3-proto.renum.pdb this is try17 on lopez hopefully these will score better than try8 and since the dimer interface of 1zp2 is larger than 1kxu's, they might be better submissions than: dimer-1kxu-try3-opt2 or dimer-1jkw-try4-opt2 Fri Jul 14 23:19:51 PDT 2006 Firas Khatib for the monomers, I think the best monomers were: try3,try4,try1,and try5. since try3 was optimized in a dimer context for both 1kxu and 1zp2, I think those two are the best: T0338.try15-opt2.unpack.gromacs0.repack-nonPC.unpack.pdb chain A and T0338.try8-opt2.unpack.gromacs0.repack-nonPC.pdb.gz chain A we also optimized try4 in a dimer context for 1kxu: T0338.try4-opt2.unpack.gromacs0.repack-nonPC.pdb.gz chain A and then the last two monomers that I think are good, but a tad different: ReadConformPDB T0338.try1-opt2.pdb ReadConformPDB T0338.try5-opt2.gromacs0.repack-nonPC.pdb try16 or try17's run might end up being better than try8, in which case we might want to replace our current 2nd best monomeric model: T0338.try8-opt2.unpack.gromacs0.repack-nonPC.pdb.gz chain A with either try16 or try17 Fri Jul 14 23:41:53 PDT 2006 Kevin Karplus I'm satisfied with Firas's choices of monomers and have submitted them, with comments: Model 1 is a monomer taken from our favorite dimer optimization (dimer/decoys/T0338.try15-opt2.unpack.gromacs0.repack-nonPC.pdb chain A) which was built from the try3-opt2 monomer. Model 2 is a monomer taken from a dimer optimization based on a different dimerization interface (modeled on the 1zp2 dimer instead of 1kxu): (dimer/decoys/T0338.try8-opt2.unpack.gromacs0.repack-nonPC.pdb.gz chain A) also optimized from the try3-opt2 monomer. Model 3 is a monomer taken from a dimer optimization based on 1kxu but built from the monomer try4-opt2, the best-scoring monomer with the unconstrained costfcn. (dimer/decoys/T0338.try4-opt2.unpack.gromacs0.repack-nonPC.pdb.gz chain A) Model 4 is try1-opt2, the model created automatically. Model 5 is try5-opt2.gromacs0.repack-nonPC, Rosetta's favorite of our backbones after repacking sidechains. ------------------------------------------------------------ Sat Jul 15 00:24:26 PDT 2006 Firas Khatib try16 & try17 are done. they both had the same dimer-interface constraints, but try17 was based off of a proteinshopped model, whereas try16 was based off try8. try16 scores best and I think should replace try8 (it doesn't have a break around L364, whereas try8 and try17 both do) in the dimer submissions our monomer could also be replaced by try16 instead of try8 for model2: before: "Model 2 is a monomer taken from a dimer optimization based on a different dimerization interface (modeled on the 1zp2 dimer instead of 1kxu): (dimer/decoys/T0338.try8-opt2.unpack.gromacs0.repack-nonPC.pdb.gz chain A) also optimized from the try3-opt2 monomer." now: "Model 2 is a monomer taken from a dimer optimization based on a different dimerization interface (modeled on the 1zp2 dimer instead of 1kxu): (dimer/decoys/T0338.try16-opt2.unpack.gromacs0.repack-nonPC.pdb.gz chain A) also optimized from the try3-opt2 monomer." I have made these 2 changes in the superimpose.under files. Sat Jul 15 01:00:11 PDT 2006 Kevin Karplus I have made the new submissions for both monomers and dimers. Note that dimer submissions depend on the dimer/Makefile, not the superimpose-best.under file, and the dimer/superimpose-best.under has the order different from the Makefile.