Tue Jul 18 09:49:03 PDT 2006 T0384 Make started Tue Jul 18 09:51:59 PDT 2006 Running on lopez.cse.ucsc.edu Tue Jul 18 14:51:09 PDT 2006 Kevin Karplus BLAST finds excellent hit for N-terminal domain (1ydwA, 27% over 145 residues, 2.6e-11). HMMs also find excellent hits (1ofgA, 1h6dA, 1ydwA, ...) possibly multi-domain hits, possibly just N-terminal domain (have to see pairwise alignments, which haven't been made yet). Tue Jul 18 15:01:17 PDT 2006 Kevin Karplus No---I don't need to wait. The .t2k alignment includes full-length PDB matches, so this seems to be a comparative model for the whole thing (unless all the multiple alignments are contaminated). Conservation gets pretty low in the second half (even for t2k), so we may want to make some subdomain predictions, to make sure we're not just seeing contamination. Tue Jul 18 22:22:14 PDT 2006 Kevin Karplus The TryAllAlign is taking a long time, as there are 5164 alignments that got converted to conformations and scwrled. If this is really fold recognition, we may be able to get away with many fewer alignments on future runs---we may even be able to do polishing without having to start again from alignments. Wed Jul 19 10:15:06 PDT 2006 Kevin Karplus All the top alignments and try1 are pretty much in agreement. There does seem to be some question about the phase of the strand Q262-P269. We may want to look for hairpin signals in the o_sep and n_sep predictions. Mon Jul 24 16:26:36 PDT 2006 Chris Wong We added some hbond constraints that we came up with from looking at o-sep predictions. # try2 increases weight: # from to # dry5 15 30 # soft_clashes 20 40 # break 50 100 # constraints 10 50 # adding constraints to fix out of phase strand Q262-P269 used o-sep Hbond N264.O D260.N 1 Hbond V266.O F258.N 1 ( make -k T0384.do2 > & do2.log ; gzip -9f do2.log ) & started on orcas at 1626. Wed Jul 26 13:26:42 PDT 2006 Chris Wong try2 scores best with unconstrained cost function, but the strands around 262-269 still don't look right. Maybe we'll need to add in some sheet constraints instead of trying to fix it using Hbond constraints. Fri Jul 28 15:43:07 PDT 2006 Pinal Kanabar I looked up o_sep and n_sep and also their rbd files - but did not find good signal for Q262-P269 try2 shorted the nice sheets which were formed with Q262-P269 may be we messed up with Hbond constraints. I talked with Chirs and we decided to give sheet constraints Following changes were made to try3.under file InfilePrefix decoys/ ReadConformPDB T0384.try1-opt2.pdb.gz include read-pdb.under try3.costfcn break 80 \ sidechain 3 \ constraints 50 \ # new : StrandConstraint S255 D260 0.80 StrandConstraint Q262 Q267 #0.80 try1-opt2.sheets constrains + //new SheetConstraint (T0384)D260 (T0384)A256 (T0384)T244 (T0384)N248 hbond (T0384)F258 10 try3 is running on camano Sun Jul 30 15:48:31 PDT 2006 Chris Wong try3 had the effect of lengthening the strand 2nd from the edge (256-259) compared with try2. Looking at the try3-opt2 model with the near script, I noticed there is another strand that is out of phase, putting hydrophobics into solution (222-228). Going to do a try4 with yet another set of sheet constraints... modified from try1-op2.sheets and .helices. ( make -k T0384.do4 > & do4.log ; gzip -9f do4.log ) & started on vashon at 439p. Sun Jul 30 23:47:42 PDT 2006 Chris Wong try4 made the strand in the area around 262-269 !! However, the other strand I was trying to change (222-228) is now a coil. I wonder why ? I'm going to do a try5 that polishes try4-op2, and maybe a try6 that basically is a try4, but without trying to fix strand 222-228. ( make -k T0384.do5 > & do5.log ; gzip -9f do5.log ) & started on orcas at 1158p. ( make -k T0384.do6 > & do6.log ; gzip -9f do6.log ) & started on orcas at Mon Jul 31 00:04:49 PDT 2006. Mon Jul 31 09:43:25 PDT 2006 Chris Wong try5. The strand that turned into a coil in try4, is now back to being a strand. However, the strand that I REALLY wanted to keep around 262-269 has been shortened and twisted. try6. Sort of the opposite af try5 happened. The strand that turned into a coil in try4 is still a coil, but the strand that I REALLY wanted is still there. How to get both nice strands? Mon Jul 31 10:36:04 PDT 2006 Pinal Kanabar with score-all+servers.unconstrained.pretty T0384.try5-opt2.pdb.gz scores best followed by T0384.try6-opt2.pdb.gz abd T0384.try3-opt2.pdb.gz grep-best-rosetta : T0378.try2-opt2.gromacs0.repack-nonPC.pdb.gz is favourit which I and Chris did not like at all - as it had shortend the strands which we were trying to fix it I am going to run polishing run on try5 and will try to get 262-269 resides as a strand. There are total 38 breaks in try5-opt2 : 5 breaks are betn 1.19 and 2.25 6 breaks are above 0.6 ... so I will try fix that by increasing wts Mon Jul 31 11:07:06 PDT 2006 Pinal Kanabar I did best-models.pdb.gz and looked at try1 try3 and try5 I dont think that it extend the strand which I was trying to extend -- will try to get that in try7 : Tue Aug 1 13:44:52 PDT 2006 Chris Wong In try7, the edge strand still looks a little weird. Also, the near script is showing two strands out of phase. Strand 196-202 and strand 222-228. Also, it looks like helix 298-317 is exposing some hydrophobics. Tue Aug 1 15:10:46 PDT 2006 Pinal Kanabar T0384.try7-opt2.pdb.gz scores best with /score-all+servers.unconstrained.pretty try7 was done with consulting George - he felt that most of the stands where not in phase. So following changes were done #try5-opt2.sheets SheetConstraint (T0384)L2 (T0384)L4 (T0384)Y26 (T0384)L28 hbond (T0384)L2 1 SheetConstraint (T0384)L4 (T0384)G8 (T0384)L65 (T0384)A69 hbond (T0384)G5 1 SheetConstraint (T0384)G5 (T0384)G8 (T0384)A30 (T0384)S33 hbond (T0384)V6 1 SheetConstraint (T0384)Q50 (T0384)L51 (T0384)Y32 (T0384)S33 hbond (T0384)Q50 1 SheetConstraint (T0384)L65 (T0384)I68 (T0384)H88 (T0384)L91 hbond (T0384)V66 1 SheetConstraint (T0384)H88 (T0384)E92 (T0384)F115 (T0384)A119 hbond (T0384)V89 1 SheetConstraint (T0384)F145 (T0384)S152 (T0384)Y239 (T0384)S232 hbond (T0384)F145 1 #SheetConstraint (T0384)D144 (T0384)S152 (T0384)Q221 (T0384)N229 hbond (T0384)N146 1 #SheetConstraint (T0384)D195 (T0384)D204 (T0384)Y217 (T0384)D208 hbond (T0384)A196 1 SheetConstraint (T0384)A196 (T0384)Y198 (T0384)E322 (T0384)R320 hbond (T0384)T197 1 #SheetConstraint (T0384)N210 (T0384)Y217 (T0384)K228 (T0384)Q221 hbond (T0384)G211 1 SheetConstraint (T0384)L234 (T0384)T240 (T0384)T249 (T0384)G243 hbond (T0384)C236 1 #SheetConstraint (T0384)244 (T0384)T249 (T0384)T259 (T0384)254 hbond (T0384)N248 1 SheetConstraint (T0384)I257 (T0384)D260 (T0384)Q267 (T0384)N264 hbond (T0384)F258 1 #new SheetConstraint (T0384)D144 (T0384)S152 (T0384)Y220 (T0384)K228 hbond (T0384)N146 10 SheetConstraint (T0384)N194 (T0384)L203 (T0384)Y217 (T0384)D208 hbond (T0384)D195 10 SheetConstraint (T0384)N210 (T0384)F216 (T0384)G227 (T0384)Q221 hbond (T0384)G211 10 SheetConstraint (T0384)T244 (T0384)T249 (T0384)T259 (T0384)R254 hbond (T0384)N248 10 #from try3.costfcn SheetConstraint (T0384)D260 (T0384)A256 (T0384)T244 (T0384)N248 hbond (T0384)F258 10 I agree with Chris that end strand is looking bad, it score best till now. And still has breaks T0384.try7-opt2.pdb.gz breaks before (T0384)S232 with cost 2.08626 T0384.try7-opt2.pdb.gz breaks before (T0384)I207 with cost 1.66144 T0384.try7-opt2.pdb.gz breaks before (T0384)Q50 with cost 1.48722 T0384.try7-opt2.pdb.gz breaks before (T0384)Y220 with cost 1.17527 T0384.try7-opt2.pdb.gz breaks before (T0384)D204 with cost 1.0211 T0384.try7-opt2.pdb.gz breaks before (T0384)Q27 with cost 0.805623 T0384.try7-opt2.pdb.gz breaks before (T0384)H275 with cost 0.702339 T0384.try7-opt2.pdb.gz breaks before (T0384)T22 with cost 0.683812 T0384.try7-opt2.pdb.gz breaks before (T0384)D64 with cost 0.597632 I have started try8 it is based on try1. It has Hbond constraints and try1-opt2.helices and sheet constriants with increased wt for break and dry wts It is running camano try8.costfcn # adding constraints to fix out of phase strand Q262-P269 # used o-sep Hbond N264.O D260.N 1 Hbond V266.O F258.N 1 #try1.sheets #wt increased to 10 #new SheetConstraint (T0384)I257 (T0384)D260 (T0384)Q267 (T0384)N264 hbond (T0384)A256 25 try8 is running on camano. Tue Aug 1 21:40:23 PDT 2006 Chris Wong try8 has made a break at the edge strand. ...terrible.. Wed Aug 2 14:29:04 PDT 2006 Chris Wong Re-run try8 as try9. The difference will be to start from alignments instead of a model. We'll use the same constraints as try8, increasing the weight of the edge strand that we want. ( make -k T0384.do9 > & do9.log ; gzip -9f do9.log ) & started Thu Aug 3 11:59:17 PDT 2006 Chris Wong try9-opt2 is really ugly at the edge strand (262-269). I'm not exactly sure why it turned out like this, but it looks like the strand is trying to form H-bonds with the strand that's sitting two positions over on the sheet. It seems like we shouldn't bother doing anything with try9-opt2, as we have other models closer to what we are going for. Pinal, do you have any ideas ? Thu Aug 3 16:34:48 PDT 2006 George Shackelford Pinal's constraints in try7 to fix the problems with big sheet look good but they didn't take. I have decided to go back to first bsae, i.e. try1 as a start and put the try7.costfcn in place of try1.costfcn. I also limited the new try to look at 1h6dA only (since that is what try1 chose). try10 running on lopez Thu Aug 3 23:07:21 PDT 2006 Pinal Kanabar try10 has got 14 breaks and 7 breaks are betn range 3.38 - 8.36 its pretty bad. try7 still scores the best with score-all+servers.unconstrained.pretty. I am not sure what else should be tried.... Fri Aug 4 09:08:33 PDT 2006 Chris Wong The current order of unconstrained cost scores for the try's is: model..............................score ========================================================================= T0384.try7-opt2.pdb.gz.............154.92 T0384.try5-opt2.pdb.gz.............156.85 T0384.try6-opt2.pdb.gz.............160.52 T0384.try3-opt2.pdb.gz.............161.19 T0384.try2-opt2.pdb.gz.............163.04 T0384.try4-opt2.pdb.gz.............165.89 T0384.try1-opt2.pdb.gz.............168.65 T0384.try8-opt2.pdb.gz.............168.95 T0384.try10-opt2.pdb.gz............172.67 T0384.try9-opt2.pdb.gz.............239.20 ========================================================================= With regards to breaks, the order is roughly the same. The edge strand (262-269) looks the best on try6-opt2. As with other models we have tried, there is a short helix near L175 that is very weakly predicted to be a strand in the concensus dssp-ehl2 logo. The t2k, t04, and t06 each have different and very weak predictions in their respective dssp-ehl2 logos, so I suppose that part of the protein could be anything. Strand at 219-228 is a little weird, too. It's weird because it's in the middle of a sheet, and predicted to be a strand, but it's a coil! I guess we should focus some attention to turning that part back into a strand. Okay, I think I'll use some constraints from try3-opt2 and the sheet constraints from try6-opt2 for the edge strand area to put together a try11 that starts with try6-opt2. ( make -k T0384.do11 > & do11.log ; gzip -9f do11.log ) & started on orcas at 1002a. Fri Aug 4 11:15:09 PDT 2006 Pinal Kanabar I dont know what else to try. I am going to start a new run from alignments with specfic Hbond and helix constraints and see if it works !!!!!!! I started new run - starting from alignments, and in cost func is similar to try7. try12 is running on Vashon. Fri Aug 4 14:41:32 PDT 2006 Chris Wong try11-opt1 has not fixed the messed up center strand. So, what I'm going to do is run a try13 that starts from try3-opt2. That was the last model that had a complete set of strands in the sheet. I'll also include the constraint for the edge strand. Hopefully, I can end up with a model that has an intact sheet plus a nice edge strand. ( make -k T0384.do13 > & do13.log ; gzip -9f do13.log ) & started on lopez at 245p. Fri Aug 4 15:22:14 PDT 2006 George Shackelford Try10 still has the main sheet screwed up; it refused to pay attention to my advice that would line up the buried residues indicated by near. I've decided to find something that might match better. I used the infamous 'alphabetmatch' with alphabets ehl2 and burial, and I got a strong hit on 1evjA. This is already in the MANUAL_TOP_HITS since it is also a strong hit in best scores. I have designed try15 to be based on 1evjA and included the t06.dssp-ehl2.constraints. I want those buried residues to line up! try15 running on peep. Fri Aug 4 15:35:03 PDT 2006 Pinal Kanabar I am going to optimize try7 as its best scoring model. My try14 is based on T0384.try7-opt2.gromacs0.repack-nonPC.pdb.gz. Goal is to try to fix end strands and reduce the breaks. try14.under : I double the cost for following parameters CloseGap 2 \ MoveGap 2 \ HealGap 4 \ ReduceBreak 4 \ cost func has constrains for Hbond and edge strands. I also increased break, soft clashes and constraints cost try14 is running on camano Fri Aug 4 20:43:06 PDT 2006 Chris Wong Here is a summary of the models from today: try11-opt2 has an edge strand, but the large sheet has a coil in it where a strand should be. It's currently 4th in unconstrained score. try12-opt2 has an edge strand, and the large sheet is mostly intact, except that there a sizeable break has been introduced. It does not score well. try13-opt2 has an edge strand that is twisted. The two strands at the edge look like they are barely held in place. They do not line up very well with the large sheet. The large sheet has a break, probably introduced in order to change the middle coil back into a strand. It scores just below try11-opt2. try14-opt1 (opt2 not available yet) has the edge strand, but it is not lined up very nicely. However, the large sheet is intact. try15-opt2 has lost that edge strand that we wanted. However, I think the large sheet has something sort of like what George was trying to go for. Fri Aug 4 21:23:51 PDT 2006 Chris Wong Okay, I just took a closer look at try13-opt2. It would appear that there are no constraints to keep the last 2 strands in the large sheet from flying away! So, the next step for me is to do a try16 that holds that sheet in place with a constraint. I'll start the try from try13-opt2, and include the edge strand constraint I used in try13.costfcn and add another constraint to keep it part of the large sheet. ( make -k T0384.do16 > & do16.log ; gzip -9f do16.log ) & started on orcas at 934p. Sat Aug 5 04:15:36 PDT 2006 Pinal Kanabar It took forever to do this with dead mac and 'Not optimized desktop' for working from home anyways... Following three are Rosetta 's faovrite T0384.try5-opt2.gromacs0.repack-nonPC.pdb.gz T0384.try7-opt2.gromacs0.repack-nonPC.pdb.gz T0384.try14-opt2.gromacs0.repack-nonPC.pdb.gz with score-all+servers.unconstrained.pretty T0384.try7-opt2.pdb.gz T0384.try7-opt1.pdb.gz T0384.try5-opt2.pdb.gz T0384.try11-opt2.pdb.gz T0384.try13-opt2.pdb.gz T0384.try7-opt2.repack-nonPC.pdb.gz T0384.try14-opt2.pdb.gz T0384.try11-opt1.pdb.gz T0384.try5-opt1.pdb.gz T0384.try5-opt2.repack-nonPC.pdb.gz I looked at try12, it is not scoring well, But I like the way it has formed first three strands (201 -222). It tore apart many things. Hey try15 also has first five sheets looks good although it completly blew up the edge strands. I looked at T0384.try16-opt1.pdb - it messed up the strand starting with 244... may its incomplete run I started : try17 : as normal run with following cost function soft_clashes 80 backbone_clashes 2 \ break 200 \ Hbond N264.O D260.N 20 Hbond V266.O F258.N 20 #from try15 SheetConstraint (T0384)M1 (T0384)L4 (T0384)E25 (T0384)L28 hbond (T0384)L220 SheetConstraint (T0384)K3 (T0384)G8 (T0384)D64 (T0384)A69 hbond (T0384)G520 SheetConstraint (T0384)L4 (T0384)I7 (T0384)V29 (T0384)Y32 hbond (T0384)V620 SheetConstraint (T0384)L65 (T0384)I68 (T0384)H88 (T0384)L91 hbond (T0384)V6620 SheetConstraint (T0384)V89 (T0384)E92 (T0384)I116 (T0384)A119 hbond (T0384)V8920 SheetConstraint (T0384)N146 (T0384)K153 (T0384)Y239 (T0384)S232 hbond (T0384)N146 20 SheetConstraint (T0384)Y147 (T0384)S152 (T0384)V222 (T0384)G227 hbond (T0384)K149 20 SheetConstraint (T0384)D195 (T0384)Q201 (T0384)F216 (T0384)N210 hbond (T0384)D195 20 SheetConstraint (T0384)N210 (T0384)L215 (T0384)G227 (T0384)V222 hbond (T0384)G211 20 SheetConstraint (T0384)P235 (T0384)T240 (T0384)N248 (T0384)G243 hbond (T0384)C236 20 # new SheetConstraint (T0384)I257 (T0384)D260 (T0384)Q267 (T0384)N264 hbond (T0384)N264 25 SheetConstraint (T0384)T246 (T0384)T249 (T0384)T259 (T0384)A256 hbond (T0384)T249 25 Aim : Hopefully undertaker will implementing sheet constraints and will come with the model which we want. try17 is running on camano I started try18 : which will optimize T0384.try12-opt2.pdb with same cost function as try18. Aim : hopefully sheet constraints will be implemented by the undertaker and reduce the breaks I started try19 : which will optimize T0384.try15-opt2.pdb with same cost function as try18. try 15 had less breaks but it completly converted edge strand into coil - so hopefully sheet constrains will bring it back. I reaaly dont know whether this is a correct path ..... Note: I am using try12 and try15 as I like the way it has formed first three strands (201 -222) although they dont score well and have terrible breaks. I am hoping that by optimizing it ...undertaker will come up with expected model. Hey Chris do you think we should email Kevin - just to look at what we have till now. May be he has some more ideas. He eamiled sombody saying he was not happy with the way target was handle... we still have whole day if he has any suggetions.....we can try few more things. We dont have anything decent unfortunatly ..... ok I am taking off it is 4.55 a.m. I have to go to stanford S M - hopefully they will fix my machine.. Sat Aug 5 07:51:39 PDT 2006 Chris Wong try14-opt2 is done. Its large sheet is intact, and it has the edge strand, but like so many of our other attempts, it just doesn't look like it's lined up very well with the rest of the sheet. try16-opt2 looks worse than try13-opt2 (the model try16 started from). It looks mostly the same as try13-opt2, but with worse break. Perhaps, it's time to select some models now and do polishing runs to remove breaks. Polishing up the following 2 models. They are 2 slightly different variations on the same edge strand (266-269). try20 will start from T0384.try12-opt2.gromacs0.repack-nonPC.pdb.gz. try12-opt2 has some serious breaks. Polishing up the gromacs-repack model should help fix that. I think try12-opt2 has the best edge strand so far. It is just slightly out of line with the rest of the sheet. Hopefully, it won't get destroyed in the polishing. try21 will start from T0384.try13-opt2.pdb.gz. This model has a little bit longer edge strand, but it is lifted off from the plane of the large sheet on one end. ( make -k T0384.do20 > & do20.log ; gzip -9f do20.log ) & started on orcas at 9a. ( make -k T0384.do21 > & do21.log ; gzip -9f do21.log ) & started on orcas at 9a. Sat Aug 5 11:00:52 PDT 2006 Pinal Kanabar try17 messed up the middle strand ... it opened up C terminal, i.e end residues which were stand are now helix; This helix is floating in air try18 was optimizing run of T0384.try12-opt2.pdb, break are still there and now it messed up last strand althoguht edge stand look better. It seems if I try to optimize first 3 strands and next three gets messed up... I reaaly dont know what to do..... try19 was based on try15 it had nice sheet constraints but the edge strand was ruined turned into coil - BUT try19 faild to recoverthose strant atleast in T0384.try19-opt1.pdb (try19 is still running..) I really dont know what five models we should submite - I like try12 and chris is optimizing it....and hopefully we will have one decent model T0384.try7-opt2.pdb.gz 154.92 T0384.try7-opt1.pdb.gz 156.67 T0384.try5-opt2.pdb.gz 156.85 T0384.try11-opt2.pdb.gz 157.00 T0384.try13-opt2.pdb.gz 158.76 T0384.try7-opt2.repack--nonPC.pdb.gz 159.31 T0384.try14-opt2.pdb.gz 159.43 T0384.try11-opt1.pdb.gz 159.84 T0384.try5-opt1.pdb.gz 159.89 T0384.try5-opt2.repack--nonPC.pdb.gz 160.18 Recap of all try's T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try2-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try3-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try4-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try5-opt2.pdb.gz -> T0384.try4-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try6-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try7-opt2.pdb.gz -> T0384.try5-opt2.pdb.gz -> T0384.try4-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try8-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try9-opt2.pdb.gz -> 1h6dA T0384.try10-opt2.pdb.gz -> 1h6dA T0384.try11-opt2.pdb.gz -> T0384.try6-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try12-opt2.pdb.gz -> 1zh8A 1rydA T0384.try13-opt2.pdb.gz -> T0384.try3-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try14-opt2.pdb.gz -> T0384.try7-opt2.gromacs0.repack-nonPC.pdb.gz T0384.try15-opt2.pdb.gz -> T0384_1evjA_long.a2m (InfilePrefix 1evjA/ ReadFragmentAlignment NOFILTER SCWRL T0384_1evjA_long.a2m) T0384.try16-opt2.pdb.gz -> T0384.try13-opt2.pdb.gz -> T0384.try3-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA T0384.try17-opt2.pdb.gz -> 1ydwA 1rydA T0384.try18-opt2.pdb.gz -> T0384.try12-opt2.pdb.gz -> 1zh8A 1rydA T0384.try19-opt2.pdb.gz -> T0384.try15-opt2.pdb.gz -> T0384_1evjA_long.a2m (InfilePrefix 1evjA/ ReadFragmentAlignment NOFILTER SCWRL T0384_1evjA_long.a2m) T0384.try20-opt2.pdb.gz -> T0384.try12-opt2.gromacs0.repack-nonPC.pdb T0384.try21-opt2.pdb.gz -> T0384.try13-opt2.pdb.gz -> T0384.try3-opt2.pdb.gz -> T0384.try1-opt2.pdb.gz -> 1h6dA We can see that maximum run are based on try1 as it had formed nice edge starnd. Most runs based on try1 were tried with different cost fuctions - but it nothing worked. Undertaker really like the models based on 1h6dA. Models based on other templetas dont even show up in top 15 best scoring model inscore-all+servers.unconstrained.pretty Following are top five Rosetta-repacked models: T0384.try5-opt2.gromacs0.repack-nonPC.pdb.gz T0384.try7-opt2.gromacs0.repack-nonPC.pdb.gz T0384.try18-opt2.gromacs0.repack-nonPC.pdb.gz T0384.try14-opt2.gromacs0.repack-nonPC.pdb.gz T0384.try8-opt2.gromacs0.repack-nonPC.pdb.gz Looking at everything : I dont feel like submitting anything.... but may be ... try7, try18 and I dont which should be thrid one. Chirsh and I agree for try12 (or optimized version based on try12- even though it doesnt score well as it has little different strand phase). He also picked up try13 (or optimized version based on try13). Lets we will have chat again see what we like... I am running out any more ideas - I will email our group and if anybody has time and if they have some suggestions..... I missed my appointment for MAC repair - so I will be home and if plp have suggetions I will be willing to implement them... I will check my email after 1 hour .. thx Sat Aug 5 13:14:44 PDT 2006 Kevin Karplus I'm doing a polishing run with high crossover from all models as try22. Please be sure to set up the superimpose-best file with the best predictions in order. Sat Aug 5 14:51:51 PDT 2006 Chris Wong Well, the try20 and try21 that I started this morning are done. Try20 (which started from try12) scores higher than try12, but try12 was pretty low to begin with (173 v. 185). Nothing much has changed, really. It still has a few bad breaks. The edge strand (262-269) is there, but sort of bent off from the plane of the large sheet it is connected to. Try21 started from try13-opt2. It is currently the top scoring model. As expected, it looks a lot like try13-opt2, which scores 7 positions below. The 2 strands at the edge are still lifted up away from the large, main sheet in the model. Sat Aug 5 15:16:45 PDT 2006 Chris Wong I just put in the following lines in the superimpose script: ReadConformPDB T0384.try21-opt2.pdb ReadConformPDB T0384.try7-opt2.pdb ReadConformPDB T0384.try18-opt2.pdb ReadConformPDB T0384.try3-opt2.pdb ReadConformPDB T0384.try10-opt2.pdb I'll do a mini-write-up after my lunch. Sat Aug 5 15:55:40 PDT 2006 Chris Wong The models are all very similar, as this is an example of comparative modelling. The try1-opt2 model had good agreement with the secondary structure prediction. The variations in the models comes from how the edge strand in the region 262-269 was formed. Model 1 is T0384.try21-opt2.pdb. This is the best scoring model by the unconstrained cost function. It comes from 3 rounds of optimizations starting from a model based on template 1h6dA. All the top alignments and try1 are pretty much in agreement, except for the the phase of the strand Q262-P269. Most of the efforts were to make the edge strand fit in with the large main sheet in the protein. This model has a nice edge strand, and good looking turn between the edge sheet and its neighbor. However, this edge strand and its neighbor are lifted up and away from the plane of the large main sheet. Model 2 is T0384.try7-opt2.pdb. This is the second top scoring model by the unconstrained cost function. This model also comes from 3 rounds of optimizations starting from a model based on 1h6dA. This model is different from try21-opt2 in that the edge strand (262-269) is a bit shorter. However, it is more lined up with the plane of the large main sheet in the protein. Model 3 is T0384.try18-opt2.pdb. The unconstrained cost score for this one is in the lower half. Though it does not score very well, it does have an edge sheet that is long like the one in try21-opt2, but more lined up with the large main sheet (like the one in try7-opt2). This model started from the templates, 1zh8A and 1rydA. (after looking at this one, going to try to reduce breaks with a try23) ( make -k T0384.do23 > & do23.log ; gzip -9f do23.log ) & started on shaw at 457p. Model 4 is T0384.try3-opt2.pdb. This model is included as a space holder for now. Model 5 is T0384.try10-opt2.pdb. This model is included as a space holder fornow. Sat Aug 5 17:53:50 PDT 2006 Pinal Kanabar Thx Kevin and Chris for taking over. Try12 was normal undertaker run with prior belifs about how the first few strands (resi : 201-195, 210-215,223-227 ) should look like. Decision was based on near script (burial). Goal was to optimze intial stands and also line up the edge stand. Chirs I am mentioning this details .. as I dont know how much details goes into method file. Sat Aug 5 19:08:41 PDT 2006 Kevin Karplus try22-opt2 has polished try7-opt2 and nudged it out of first place with the try22 costfcn. It has also moved up to 2nd place with the rosetta scoring (aas try22-opt2.gromacs0.repack-nonPC.pdb.gz). Based on the notes above, I would rank the models try22-opt2 # best with try22 (high clash unconstrained) try20-opt2.gromacs0.repack-nonPC #rosetta favorite try21-opt2 try18-opt2 try3-opt2 I have not looked at them yet (as download is very slow tonight). Sat Aug 5 20:49:57 PDT 2006 Kevin Karplus try3-opt2 is too similar to try22-opt2, so I replaced it with try1-opt2, which has teh distinction of being the automatic model. I have submitted predictions, but if try23 finishes and looks good, it can replace try18-opt2. Submitted with comment: The models are all very similar, as this is an example of comparative modelling. The automatic try1-opt2 model had good agreement with the secondary structure prediction and good overall fit. The variations in the models comes from how the edge strand in the region Q262-P269 was formed. Model 1 is try22-opt2. It is the best-scoring model with undertaker unconstrained cost functions. It was optimized by undertaker from try7-opt2, which came from three rounds of optimization by undertaker of an alignment to 1h6dA. Model 2 is try20-opt2.gromacs0.repack-nonPC, which is try20-opt2 repotimized by gromacs and with sidechains repacked by rosetta. It was rosetta's favorite of the backbons it repacked. Model 3 is T0384.try21-opt2.pdb. This is the best scoring model by the unconstrained cost function (after try22-opt2). It comes from 3 rounds of optimizations starting from a model based on template 1h6dA. All the top alignments and try1 are pretty much in agreement, except for the the phase of the strand Q262-P269. Most of the efforts were to make the edge strand fit in with the large main sheet in the protein. This model has a nice edge strand, and good looking turn between the edge sheet and its neighbor. However, this edge strand and its neighbor are lifted up and away from the plane of the large main sheet. Model 4 is T0384.try18-opt2.pdb. The unconstrained cost score for this one is in the lower half, because of bad breaks. Though it does not score very well, it does have an edge sheet that is long like the one in try21-opt2, but more lined up with the large main sheet (like the one in try2-opt2). This model started from the templates, 1zh8A and 1rydA. Model 5 is try1-opt2, the fully automatic model. ---------------------------------------------------------------------- Sat Aug 5 21:42:24 PDT 2006 Chris Wong try23-opt2 has still got some bad breaks. However, it is #51 on the list of undertaker unconstrained scores where try18-opt2 is #91. Try23 was started from try18-opt2. Looking in grep-best-rosetta, T0384.try23-opt2.gromacs0.repack-nonPC.pdb was #5 on the list while T0384.try18-opt2.gromacs0.repack-nonPC.pdb was #6. It's virtually a tie. The models themselves look very similar. I'm thinking try23-opt2 should replace try18-opt2 as model 4 in the submissions. Sat Aug 5 22:10:46 PDT 2006 Chris Wong I've changed the superimpose-best.under to this: ReadConformPDB T0384.try22-opt2.pdb ReadConformPDB T0384.try20-opt2.gromacs0.repack-nonPC.pdb ReadConformPDB T0384.try21-opt2.pdb #ReadConformPDB T0384.try18-opt2.pdb ReadConformPDB T0384.try23-opt2.pdb ReadConformPDB T0384.try1-opt2.pdb T0384.method is changed to have this: Model 4 is T0384.try23-opt2.pdb. The unconstrained cost score for this one is in the middle third of the list, because of bad breaks. Though it does not score very well, it does have an edge sheet that is long like the one in try21-opt2, but more lined up with the large main sheet (like the one in try2-opt2). This model started from the templates, 1zh8A and 1rydA. Sat Aug 5 22:23:30 PDT 2006 Chris Wong Resubmitted with model 4 as try23-opt2 with the following steps: 1) edit superimpose-best.under 2) edit T0384.method 3) make best-models.pdb.gz 4) make casp_models 5) make email Sat Aug 5 22:30:17 PDT 2006 Chris Wong I've just confirmed at the model list viewer website that the resubmission has been accepted. T0384TS010_1 08/05/06 22:21:03 T0384TS010_2 08/05/06 22:21:15 T0384TS010_3 08/05/06 22:21:26 T0384TS010_4 08/05/06 22:21:37 T0384TS010_5 08/05/06 22:21:48 T0384RR010_1 08/04/06 15:39:55 Sun Sep 24 14:33:12 PDT 2006 Kevin Karplus For this target HHpred3 and ROBETTA did well: HHpred3_TS1 -104.51 ROBETTA_TS1 -101.49 our best submitted model4 (try23-opt2) -52.44 our best generated try23-opt2.gromacs0.repack-nonPC -55.96 None of our servers did well: SAM-T99_AL5-scwrl -22.08 SAM_T06_server_TS4 -8.74 SAM_T06_server_TS1 1.52 We never even *looked* at the Robetta models for this target! sigh.