Wed Jun 21 10:32:20 PDT 2006 T0339 Make started Wed Jun 21 10:33:25 PDT 2006 Running on lopez.cse.ucsc.edu Wed Jun 21 10:41:55 PDT 2006 Kevin Karplus BLAST gets excellent full-length hits in PDB: 1p3wA 31% over 416 residues 7.9e-48 1eg5A 32% over 416 residues 5.1e-47 Wed Jun 21 19:20:35 PDT 2006 Kevin Karplus The HMMS get lots of hits to SCOP domain c.67.1.3 (which includes the two excellent BLAST hits). There are also at least 120 c.67.1.* hits. Thu Jul 6 09:49:23 PDT 2006 Pinal Kanabar I will follow other dimer readme files for this target. I looked at the summary.html : T0399.best_hits : 1eg5A (8.9495e-37) and 1p3wA(3.2352e-30) are ranked 3rd and 5th by undertaker. All the top 10 hits have same scope domian c.67.1.3 (Class: Alpha and beta proteins (a/b) Mainly parallel beta sheets (beta-alpha-beta units)) Thu Jul 6 10:57:51 PDT 2006 Pinal Kanabar I used near script and spacefilled option to look the protein and it seems one whole side wants to be buried but it is exposed and it seems this the places where other unit might bind. Checked for the conserved residues and conserved_t04 and conserved_t06 agrees with each other. T0339.try1-opt2.pdb.gz, 1eg5.pdb and 1p3w all has mixture of parallel and anti-parallel beta-sheets. I think I need to talk to Grant before proceeding - Anyways I followed the instructions for dimer and created dimer directory . I am stuck at make-dimer.under. Sun Jul 9 13:25:12 PDT 2006 I have started residue comparistion : Top 65 resi fold seems to be in agreement with 1p3w. Sun Jul 9 14:36:53 PDT 2006 T0339 Protein length : 416 Select 1-10 : I get 89 atoms (T0339.try1-opt2.pdb.gz) where asfor 1p3w I get 152 atoms -- I dont if this means some atoms are missing. RESIDUE ANLYSIS Residues 1-10 try1-opt2 : 89 atoms :very small beta strand and beta ...(?) VAL 4 and TRY 5 forms parrallel B- strand with LEU-352 and MET 353 1p3w : 152 atoms : In chain A : ILE 5 and TRY 6 forms parallel beta sheet with LEU 320 and ALA 321 Note: try1 has extra beta-bulge(?). Residue 15-30 : try1: helix (prdiction matches : dssp-ehle2) 1p3w : 15-30 helix 1eg5 : got break in bet arount 23 rd resi - i dont understand So will keep that heliex. Residue : 40-50 helix : in agreement with 1p3w Residue : 70-85 : helix almost all agree. Residue : 85-90 : try1 and 1eg5 : coil, 1p3w:beta strand : 96-100 : beta strand 1p3w and leg5 : TRY1 COIL This method is taking really long with rasmol being slow. From T0339.dssp-ehl2 I am going to go over the portions which have week predictions.... Mon Jul 10 09:45:25 PDT 2006 Grant Thiltgen I'm going to start a couple of polishing runs on the model we have. It seems to match the alignments very well, with a few loops that have some fixing up to do. Try2 is a polishing run with dry6.5 at 25, dry8 at 20, phobic_fit at 3, break at 100, soft_clashes at 40, hbond_geom at 10, pred_alpha04 at 3, and pred_alpha04 at 4. Try3 is like try2, but with dry6.5 at 30, dry8 at 25, soft_clashes at 60, and break at 150. Both of them are polishing try1-opt2. I'm hoping to clean up some of the gaps in the model before dimerizing it. Try2 and try3 started on orcas. Mon Jul 10 17:19:45 PDT 2006 Grant Thiltgen Try4 is going to polish try2. It scores slightly better than try3 in the unconstrained costfcn. I raised soft_clashes to 80, break to 200, hbond_geom to 20, dry6.5 to 35 and dry8 to 30. Try4 started on vashon. Tue Jul 11 09:11:02 PDT 2006 Grant Thiltgen I'm not sure try4 is all that better than try2. I'm going to go ahead and start a dimer based on try4 in order to get a dimer ready for the soft deadline on Wednesday. I'm going to dimerize try4 and optimize it some. I'm still going to work on optimizing the monomer to find a better dimer for Friday's deadline. Tue Jul 11 11:30:40 PDT 2006 Grant Thiltgen I dimerized try4 using the 1t3iA template, as it scored the best. I added all the multimer constraints to the try1.costfcn and try1.under files. I reduced the amount of iterations, since it is kind of a big protein. I am starting try1 for the dimer on orcas. Tue Jul 11 12:32:24 PDT 2006 Grant Thiltgen I also noticed that the top alignment wasn't for the model that was the top pdb hit. I'm wondering if we're modelling too much on some of the other distant homologs, and not on the direct ones. I'm going to do try5 from the beginning, using only the two models that are very close in the PDB to see if we may get a better structure for the monomer. I'm running try5 of the monomer using the structures 1p3wA and 1eg5A try5 started on vashon. Tue Jul 11 15:42:22 PDT 2006 Kevin Karplus try5 is rather risky, as it has no constraints on the model. It might be better to include the dssp-ehl2 constraints, and the sheet constraints from alignments to 1p3wA and 1eg5A. try5 appears to have picked up a global alignment to 1eg5A (though it almost went with a 1pmmA alignment from the included alignments---why were they included if you wanted only 1p3wA and 1eg5A to be considered?). This is likely to be a relatively poor alignment to 1eg5A. Ah--I see, you had no already constructed sheet constraints to use for these two templates. Doing InfilePrefix 1eg5A/ ReadFragmentAlignment NOFILTER SCWRL T0339-1eg5A-t2k-local-str2+near-backbone-11-0.8+0.6+0.8-adpstyle5.a2m InfilePrefix 1p3wA/ ReadFragmentAlignment NOFILTER SCWRL T0339-1p3wA-t2k-local-str2+near-backbone-11-0.8+0.6+0.8-adpstyle5.a2m PrintAlignmentsSheets T0339.1eg5A-1p3wA.sheets would generate the sheet constraints from the usually best local alignment. (You need all the usual startup stuff for undertaker---see starter-directory/show-align.under) Tue Jul 11 16:35:21 PDT 2006 Grant Thiltgen I made the T0339.1eg5A-1p3wA.sheets. I am going to run try6. Try6 will have the sheet constraints from the T0339.1eg5A-1p3wA.sheets. I also included the dssp-ehl2.constraints. I'm not sure if it will be done by 8 pm, but it might be ready for tomorrow. It will be ready for the hard deadline on Friday. Try6 started on whidbey. I am also hoping to get try5 dimerized in time for submission, but with such a large protein, I don't have much hope for it to be done by 8 pm. I should have at least one dimer for submission tonight. Tue Jul 11 17:44:37 PDT 2006 Kevin Karplus Note: there is a discrepancy between the 1eg5A alignment and the 1p3wA alignment: SheetConstraint G310 R313 I334 N331 hbond Q311 1 SheetConstraint G310 I314 R335 N331 hbond K312 1 It might be worthwhile to decide which of the alignments is preferable and keep just that constraint. Tue Jul 11 17:53:14 PDT 2006 Grant Thiltgen I just dimerized try5-opt2 using 1p3w as a template. I started the polish of that (try2) on shaw. This is just for preliminary submissions though, I am also going to polish try5 by itself to get a better monomer for dimerization. Tue Jul 11 18:02:36 PDT 2006 Grant Thiltgen I started try7, a polish of try5-opt2. It uses the same costfcn as try2 for a starting polishing run. Try7 started on whidbey. Tue Jul 11 18:28:52 PDT 2006 Grant Thiltgen Currently, I have the following in for submission for the monomer: T0339.try4-opt2.pdb T0339.try5-opt2.pdb T0339.undertaker-align.pdb model 1 T0339.undertaker-align.pdb model 2 T0339.undertaker-align.pdb model 3 T0339.try4-opt2 is a polish of the original undertaker model. T0339.try5-opt2 was my first attempt to use only the two pdb models that were the top scoring (I may have screwed that up, but I have a model) The rest are alignments. I have more stuff running. I'd like to submit a monomer of the try1-opt2 dimer that is currently running, unfortunately it is still running and I can't do anything until it's finished, but I'm not going to be available for awhile after 7:00 and I'm not sure it will be finished by then. I also have try6 running which is a better attempt at using only the two pdb models, and try7 which is a polish of try5. Try6 and try7 are unlikely to be finished by 8 pm. Current, I have the following in for submission for the dimer: T0339.try1-opt2.pdb dimer-1t3iA-try4-opt2.pdb dimer-1p3wA-try5-opt2.pdb T0339.try1-opt2.pdb is actually still running. I'm hoping it will be finished by 8:00, but I don't know. It is a polish of dimer-1t3iA-try4- opt2.pdb. The other two models are just the original dimers made from the templates and the two monomers. There are likely to be clashes. I'm running a polish on dimer-1p3wA-try5-opt2.pdb which is try2 in the dimer directory, but it is very unlikely to be finished by 8 pm so I haven't included it in the superimpose-best.under file for submission. I hope these will be finished sometime late tonight, and I can update the submissions to be re-submitted in the morning, but by 8 pm, this is all I have. Tue Jul 11 21:10:05 PDT 2006 Kevin Karplus try6 is still done without constraints, because the atom names are wrong in the constraints: # SheetConstraint Error: residue specified as (T0339)D63 doesn't match (T0339)Q63 # Error: can't parse residue name in position0 ... I think the problem was that Grant did not replace the place filler START_COL with 1 in his undertaker script, and further did not do a make decoys/score-all.try6.pretty to verify that the sheets constraints were correct. I moved his script to make-sheets.under-old and fixed the script in make-sheets.under I created try8.costfcn the same as try6.costfcn, but with the corrected sheet constraints. Tue Jul 11 21:21:23 PDT 2006 Kevin Karplus try8 started on cheep (like try6, but with corrected sheet constraints). The try8 costfcn actually likes try6-opt2 best (at the moment), so it is still worth considering. Rosetta likes try6-opt2.gromacs0.repack-nonPC best. There is a terrible gap before P370 in try6-opt2, and the gromacs model really messes things up to smear out the break. The loop region G357-R387 needs attention, as does N89-P108. I think I'll submit try4-opt2 try6-opt2.gromacs0.repack-nonPC (Rosetta favorite) try5-opt2 align1 (1t3iA) align2 (1v2dA) as the preliminary submission. We may need to do more sampling of the big loop regions, or hope that polishing will reshape the loops into something more convincing. Tue Jul 11 21:53:38 PDT 2006 Kevin Karplus Preliminary submission done, with comments: For the preliminary submission: Model 1 is try4-opt2, the best-scoring model with an unconstrained cost function. Model 2 is try6-opt2.gromacs0.repack-nonPC, Rosetta's favorite for repacking sidechains, which still has terrible damage done by gromacs in trying to close a gap near P370. try6-opt2 is based mainly on 1p3wA. Model 3 is try5-opt2, which is based mainly on 1eg5A. Model 4 is sidechain replacement by SCWRL on an alignment to 1t3iA. Model 5 is sidechain replacement by SCWRL on an alignment to 1v2dA. Several of the big loops have not been closed yet. Now I have to do the dimer submission. Tue Jul 11 22:07:36 PDT 2006 Kevin Karplus Preliminary dimer submission done with comment Dimer 1 is try1-opt2 our first optimization in dimer context for T0339. Dimer 2 is try1-opt2.unpack.gromacs0.repack-nonPC, the same model after optimization with gromacs and rotamer optimization by rosetta. Dimer 3 is dimer-1t3iA-try4-opt2, the monomer try4-opt2 superimposed on the 1t3i dimer with no optimization in dimer context. Dimer 4 is dimer-1p3wA-try5-opt2, the monomer try5-opt2 superimposed on the 1p3w dimer with no optimization in dimer context. Wed Jul 12 09:44:28 PDT 2006 Grant Thiltgen After everything has run, try4 still does best on the unconstrainted costfcn. Try7, which is a polish of try5 is next. Then try6 and try8 (which are very close). I'm going to start some polishing runs of try6 and try8. Also another polish of try7. I hope to get some good models by today or tomorrow to make some better dimers with for submission. Try9 which is a polish of try7 is started on whidbey. Try10 which is a polish of try6 is also started on whidbey. Try11 which is a polish of try8 is started on orcas. Wed Jul 12 17:18:42 PDT 2006 Grant Thiltgen After polishing, the top scoring model on the unconstrained costfcn is try10 which is a polish of try6. Try11 scores next which is a polish of try8. Try9 scores slightly better than try7 which is the model it is polishing. I am going to run at least one more polishing run on these before attempting to dimerize them. Try12 is a polish of try9-opt2.gromacs0 since the polish seemed to have no effect. Try13 is a polish of try10-opt2. Try14 is a polish of try11-opt2. Try12 started on whidbey. Try13 started on camano. Try14 started on orcas. Thu Jul 13 12:24:22 PDT 2006 Grant Thiltgen After polishing the monomers, the unconstrained costfcn has try13 as the best, then try10, try14, try4, try11, try2, try3, try9. Since some are polish models, the current ranking for the monomers are: Try13-opt2 Try14-opt2 Try4-opt2 Try9-opt2 Since this is comparative modeling, I'm sure that they four models are all very similar. I'm updating the superimpose-best.under file to reflect these changes. I am also going to work on a dimer of try13, try12, and try9 since these haven't been dimerized yet. I am also going to start a new monomer run that starts from the beginning with all the alignments to see if we can get something slightly different than try4. Thu Jul 13 13:31:52 PDT 2006 Grant Thiltgen I made dimers of try13, try14, and try9. For try13 and try14 I used 1p3wA as a template for dimerization and for try9 I used 1eg5A. I am running the dimers through undertaker for optimization. try3 is on the dimer for try14, try4 is for the dimer of try13 and try5 is for the dimer of try9. dimer/try3 started on orcas. dimer/try4 started on camano. dimer/try5 started on whidbey. I made try15 for the monomer, which is just starting at the beginning of the run again with all the alignments, just to see if something different than try1 turns up. I did turn off the rr.constraints for try15. Try15 started on vashon. Thu Jul 13 22:52:05 PDT 2006 Grant Thiltgen Try15 doesn't look too good, but I'll try polishing it up anyway. Try16 will be a polish of try15. Try16 started on shaw. For the dimers, the current best model by the unconstrained costfcn is try4. Try5 didn't dimerize well and there were too many clashes so undertaker tore them apart. I am going to polish up try1-try4. Try6 is a polish of try1, try7 is a polish of try2, try8 is a polish of try3, try9 is a polish of try4. Try6 started on shaw. Try7 started on lopez. Try8 started on lopez. Try9 started on vashon. Fri Jul 14 12:36:30 PDT 2006 Grant Thiltgen For the dimer, the current best scores in the unconstrained costfcn are Try9-opt2 (which is a polish of try4 and a dimer of the try13 monomer) Try4-opt2 Try8-opt2 (which is a polish of try3 and a dimer of the try14 monomer) Try6-opt2 (which is a polish of try1 and a dimer of the try4 monomer) Try3-opt2 Since two of these are based on the same monomer, I am going to look at some of the others to see how well they did. The all seem to overlap pretty consistantly with a few loops on the outside of the protein varying. I decided on this order for the dimers: T0339.try6-opt2.pdb T0339.try9-opt2.pdb T0339.try8-opt2.pdb T0339.try7-opt2.pdb dimer-1t3iA-try4-opt2.pdb T0339.try6-opt2.pdb is a polish of try1 which is the dimer of the try4 monomer. I decided to put this one first, because try4 was the final polish of the first undertaker run, and it is likely to be better than anything I did by hand for this comparative modeling protein. T0339.try9-opt2.pdb is the top scoring dimer and is a polish of the try4 dimer. The try4 dimer is a simer of the try13 monomer. I put this next because it was top scoring. T0339.try8-opt2.pdb is the next monomer for scoring. It is a polish of try3, which is a dimer of the try14 monomer. T0339.try7-opt2.pdb is a polish of try2, which is a dimer of try8. dimer-1t3iA-try4-opt2.pdb is the unpolished dimer of the try4 monomer using 1t3iA as a template. For the monomer, I have these models: T0339.try4-opt2.pdb T0339.try14-opt2.pdb T0339.try13-opt2.pdb T0339.try9-opt2.pdb I'm not sure what to include for the last model. Either try1 because it was the undertaker run or one of the alignments. T0339.try4-opt2.pdb is a polished model of try1, which is the original undertaker run. It ended up being based mostly on 1t3iA T0339.try14-opt2.pdb is a polish of try8. Try8 is a model attempting to use the top two pdb hits, and with the proper sheet constraints so that it would run with constraints as opposed to without them. It does score slightly worse on the unconstrained costfcn, but it works better with the dimerization. T0339.try13-opt2.pdb is a polish of try6, which ended up being a model that I tried to start with just the alignments from the top two pdb hits, but I ended up running without constraints. It ended up being mostly 1p3wA. T0339.try9-opt2.pdb is a polish of try5, which was an attempt to run off of just the top two pdb hits, but I ran without constraints. It ended up being mostly 1eg5A. I have also edited the dimer files to only have the monomer to see how they compare to the originals. If they do we well, we may want to include them instead of or in addition to the current ones. Fri Jul 14 15:31:17 PDT 2006 Grant Thiltgen I'm running one more polishing run on the "best" rosetta model, T0339.try13-opt2.gromacs0.repack-nonPC.pdb.gz. I increased break to 400, bad_peptide to 30, and soft_clashes to 160. This is try17, and started on whidbey. Fri Jul 14 18:01:01 PDT 2006 Grant Thiltgen Try17 is finished. It scores worse than try13 on the undertaker unconstrained costfcn, but it scores better on Rosetta. I'm going to make a dimer and try to optimize it, just to see. I am making it a dimer using 1p3wA as a template. I am running it as try10 in the dimer directory. Try10 is started on vashon. Fri Jul 14 20:27:48 PDT 2006 Grant Thiltgen Currently, the best for the monomers are T0339.try4-opt2.pdb T0339.try14-opt2.pdb T0339.try13-opt2.pdb T0339.try9-opt2.pdb T0339.try17-opt2.gromacs0.repack-nonPC.pdb T0339.try4-opt2.pdb is a polished model of try1, which is the original undertaker run. It ended up being based mostly on 1t3iA T0339.try14-opt2.pdb is a polish of try8. Try8 is a model attempting to use the top two pdb hits, and with the proper sheet constraints so that it would run with constraints as opposed to without them. It does score slightly worse on the unconstrained costfcn, but it works better with the dimerization. T0339.try13-opt2.pdb is a polish of try6, which ended up being a model that I tried to start with just the alignments from the top two pdb hits, but I ended up running without constraints. It ended up being mostly 1p3wA. T0339.try9-opt2.pdb is a polish of try5, which was an attempt to run off of just the top two pdb hits, but I ran without constraints. It ended up being mostly 1eg5A. T0339.try17-opt2.gromacs0.repack-nonPC.pdb is a polish of the best model of the rosetta score function. It was T0339.try13-opt2.gromacs0.repack-nonPC. I polished it with undertaker and now the score is better for this model so I'm including it here. For the dimer I have: T0339.try6-opt2.pdb T0339.try9-opt2.pdb T0339.try8-opt2.pdb T0339.try7-opt2.pdb dimer-1t3iA-try4-opt2.pdb T0339.try6-opt2.pdb is a polish of try1 which is the dimer of the try4 monomer. I decided to put this one first, because try4 was the final polish of the first undertaker run, and it is likely to be better than anything I did by hand for this comparative modeling protein. T0339.try9-opt2.pdb is the top scoring dimer and is a polish of the try4 dimer. The try4 dimer is a simer of the try13 monomer. I put this next because it was top scoring. T0339.try8-opt2.pdb is the next monomer for scoring. It is a polish of try3, which is a dimer of the try14 monomer. T0339.try7-opt2.pdb is a polish of try2, which is a dimer of try8. dimer-1t3iA-try17-opt2.pdb is the unpolished dimer of the try4 monomer using 1t3iA as a template. I also have a dimer of T0339.try17-opt2.pdb running, although I didn't make one of the gromacs version. If you want to include this when it's finished it's fine, or you can just include dimer-1p3wA-try17-opt2.pdb which is the dimer without any optimization. Fri Jul 14 22:45:06 PDT 2006 Kevin Karplus I wonder why Grant put the try17-opt2 run 5th on his list, when it scores the best in rosetta and with the try17 costfcn. He also hasn't listed the monomers from the dimers, some of which score better with the unconstrained costfcn. (monomer-from-try9-opt2-dimer, monomer-from-try6-opt2-dimer, monomer-from-try8-dimer are all up with try13-opt2) Why is try4 Grant's favorite monomer??? Fri Jul 14 23:54:11 PDT 2006 Grant Thiltgen Sorry about that, I thought I had made some comments about some of the choices, but I made them on an earlier comment about top models earlier in the day. I chose try4 because it was the polish from the first run from undertaker. Since it is a comparative model, I was concerned that attempting to change things (like using specific fragments from proteins) may have actually made things worse instead of better even though it scored better in the cost functions. If this isn't a problem, and choosing the structures that score better would be better even if we decided to mess around with the comparative model, then moving one of the better scoring models to the top model really isn't a problem. I was also concerned about if optimizing as the dimer might screw things up for comparative modeling too. Again, if it's not a problem, choosing the monomers from the optimized dimer for some or all of the runs really isn't a problem for me either. Since the monomer from the try9 dimer, the try6 dimer and the try8 dimer scored better, I can sub these out. The try9 dimer was from try13, the try6 dimer is from try4, and the try8 dimer was from try14. The try7 dimer was from the try8 monomer, and scored worse as a monomer from the dimer than the monomer alone. In that case, the monomer submission should be: T0339.monomer-from-try9-opt2-dimer.pdb T0339.monomer-from-try6-opt2-dimer.pdb T0339.try17-opt2.gromacs0.repack-nonPC.pdb T0339.monomer-from-try8-opt2-dimer.pdb T0339.try9-opt2.pdb T0339.monomer-from-try9-opt2-dimer.pdb is the monomer from the try9-opt2 dimer. This dimer came originally from try13, which was ultimately a polish of try6 which ended up being a model I tried to base on the top two PDB hits. It ended up being mostly 1p3wA. I messed up the constraints for this model, so it probably ran without constraints, but it still scores well. T0339.monomer-from-try6-opt2-dimer.pdb is the monomer from the try6-opt2 dimer. This dimer came originally from try4, which was ultimately a polish of try1, which was the original undertaker run. This could be our best result because it's unlikely that I could improve over the comparative model. T0339.try17-opt2.gromacs0.repack-nonPC.pdb is also a polish of try13, but as a monomer. It used try13-opt2.gromacs0.repack-nonPC.pdb to polish from since this was the model that rosetta liked the best. I polished it, and rosetta liked this repack model the best after polishing, so I'm including it here. T0339.monomer-from-try8-opt2-dimer.pdb is the monomer from the try8-opt2 dimer. This dimer came originally from try14, which was ultimately a polish of try8, which was also a model trying to be based on the top two PDB hits, but it was run with sheet constraints from the two alignments. T0339.try9-opt2.pdb is a polish of try5, which was an attempt to run off of just the top two pdb hits, but I ran without constraints. It ended up being mostly 1eg5A. This model scored better as the monomer than it did with dimerization. The dimer submission should be: T0339.try9-opt2.pdb T0339.try6-opt2.pdb T0339.try8-opt2.pdb T0339.try10-opt2.pdb T0339.try7-opt2.pdb T0339.try9-opt2.pdb is the top scoring dimer and is a polish of the try4 dimer. The try4 dimer is a simer of the try13 monomer. I put this next because it was top scoring. T0339.try6-opt2.pdb is a polish of try1 which is the dimer of the try4 monomer. This is based off the original undertaker results from the first run of undertaker. This could be our best model just because it's unlikely I could improve a comparative modeling result. T0339.try8-opt2.pdb is the next monomer for scoring. It is a polish of try3, which is a dimer of the try14 monomer. Even though it scored slightly better than try6, I chose it next because the monomer from try6 scores slightly better. T0339.try10-opt2.pdb is based on an optimization from the best scoring rosetta hit, try17. T0339.try7-opt2.pdb is a polish of try2, which is a dimer of try8. Sat Jul 15 00:29:48 PDT 2006 Kevin Karplus Monomer models submitted in the order Grant suggested. If we messed them up in the optimization, we'll find out about it! There were some big gaps to close, so I suspect we did improve things. I'm changing the dimer submission to T0339.try9-opt2.pdb T0339.try10-opt2.unpack.gromacs0.repack-nonPC.pdb Rosetta favorite T0339.try6-opt2.pdb T0339.try8-opt2.pdb T0339.try7-opt2.pdb Sat Jul 15 00:50:22 PDT 2006 Kevin Karplus dimer submission done.