Fri Jul 14 10:50:02 PDT 2006 T0378 Make started Fri Jul 14 10:52:12 PDT 2006 Running on lopez.cse.ucsc.edu Fri Jul 14 15:39:28 PDT 2006 Kevin Karplus BLAST has excellent full-length hit to 1ipaA (26% on 265 residues E-value 1.6e-15), 1x7oA, and 1zjrA. HMMs seem to be finding same stuff (not finished yet). Fri Jul 14 18:14:44 PDT 2006 Kevin Karplus The HMMs favor 1gz0A, 1x7oA, 1ipaA, 1v2xA, ... try1 seems to be based on 1zjrA (after first preferring 1x7oA). Wed Jul 26 15:06:04 PDT 2006 George Shackelford I assisted Pinal in starting up a polishing of try1. Pinal is starting try2 on ... cheep! Fri Jul 28 11:24:57 PDT 2006 Pinal Kanabar I run try3 yesterday but forgot to put comment in the readme file try3 was same as try2 with increased wt on break and sof clases and dry. In try3 there is one really big break in try4 I will try4 seal that break following changes were made to try4.costfcn #try3-opt2.helices #changed : hopeing that this will be able to close the break #HelixConstraint (T0378)E68 (T0378)R74 #to #HelixConstraint (T0378)S69 (T0378)K73 #and added #HelixConstraint (T0378)A49 (T0378)T57 0.8 changed: sidechain 5 \ -> sidechain 3 \ Mon Jul 31 09:42:42 PDT 2006 Pinal Kanabar In try 4 also there is big break between resi. PRO 81 and GLN 82 It is little strange : like when I select this residues with following command select 81-82 color blue nothing happens but when I switch the smae residues to sapcefill mode ... they are nice packed .... I dont understand why they are invisible in cartoon mode try4 (15 breaks) has fewer break than try1-opt2 (25 breaks) and also cost has been reduced. Conformation[21] T0378.try4-opt1.pdb.gz has 15 breaks T0378.try4-opt1.pdb.gz breaks before (T0378)Q82 with cost 0.864888 T0378.try4-opt1.pdb.gz breaks before (T0378)N238 with cost 0.328627 T0378.try4-opt1.pdb.gz breaks before (T0378)G149 with cost 0.324432 T0378.try4-opt1.pdb.gz breaks before (T0378)L84 with cost 0.295158 I will try to do more polishing on try4 and see if this break get redueced... Mon Jul 31 12:58:14 PDT 2006 George Shackelford I've looked at Pinal's break and I don't see a straight-forward solution. I don't really see why the break is even there except the conformation is so difficult. Perhaps if we just try compacting the structure (it looks foamy) the gap will close on its own. Tue Aug 1 13:35:43 PDT 2006 George Shackelford I have started a new polishing run using try4-opt2.gromacs0.repack-nonPC. I have break at 300 and soft clashes at 50(?) try5 running on vashon. (I see that the breaks have gone to zero!) Tue Aug 1 16:02:59 PDT 2006 Pinal Kanabar I looked at try5 and I dont know its all spreaded --- it looks very differernt from try4 and try1 Tue Aug 1 17:24:18 PDT 2006 George Shackelford Ouch. Ugly horrible things like try5 can happen when you crank up soft clashes and breaks too high. It solves the problem by blowing up the structure. See? No clashes. No breaks. No form! Ok, we'll give it another try but turn down the settings: breaks to 150 (my fav) and soft clashes to 40. try6 running on peep. Wed Aug 2 09:15:05 PDT 2006 Pinal Kanabar I dont know but try6 also looks weird. I was looking at try5.under I found that : InfilePrefix decoys/ // include read-pdb.under ReadConformPDB T0378.try5-opt2.gromacs0.repack-nonPC.pdb CostConform I dont know how can we have T0378.try5-opt2.gromacs0.repack-nonPC.pdb before running the try5... I hoping that this mistak - which happened while replacing all try4 with try5... I am going to start new run (try7) with T0378.try4-opt2.gromacs0.repack-nonPC.pdb and see what happenes. try7.under : InfilePrefix decoys/ ReadConformPDB T0378.try4-opt2.gromacs0.repack-nonPC.pdb try7.costfcn SetCost wet6.5 15 near_backbone 5 way_back 5 dry5 20 dry6.5 40 dry8 30 dry12 8 \soft_clashes 70 backbone_clashes 2 \ break 250 \ try7 is running on shaw Wed Aug 2 10:04:09 PDT 2006 George Shackelford The appearance fo teh gromacs0.repack file appearing before the run of try5 is actually due to an aborted run. When a run falters due to a bug or whatever, that particular file is usually created. It is an empty file however. Don't worry about it. These failures are discouraging. We should try this using the standard try4-opt2.pdb instead of the gromacs.repack. as a starting point. Again we need to carefully limit the weight on soft clashes and breaks. Wed Aug 2 10:37:43 PDT 2006 Pinal Kanabar I guess I was not clear - we can not start try5 with T0378.try5-opt2.gromacs0.repack-nonPC.pdb right because we are just starting try5(try5.under) .. WE DONT HAVE (T0378.try5-opt2.gromacs0.repack-nonPC). Am I right? Or may be I dont that bug changes try5.under file........ Anyways after you come to school we can start some more runs. I am running try7 with T0378.try4-opt2.gromacs0.repack-nonPC and it should be done soon... Wed Aug 2 15:47:13 PDT 2006 Pinal Kanabar I looked at try7 and lookes similar to our previous models. With score-all+servers.unconstrained.pretty try3 and try4 score best - it means increasing break cost or starting from gromacs is not helping. T0378.try3-opt2.pdb T0378.try4-opt2.pdb.gz T0378.try4-opt1.pdb.gz T0378.try3-opt1.pdb T0378.try3-opt2.pdb.gz T0378.try3-opt1.pdb.gz T0378.try3-opt2.repack-n T0378.try4-opt2.repack-n T0378.try7-opt2.pdb.gz T0378.try2-opt2.pdb.gz I started try8 with different alignments - HMM favorite and some of the top PDB hits InfilePrefix 1gz0A/ include read-alignments-scwrl.under InfilePrefix 1v2xA/ include read-alignments-scwrl.under InfilePrefix 1ipaA/ include read-alignments-scwrl.under InfilePrefix 1mxiA/ include read-alignments-scwrl.under InfilePrefix 1z85A/ include read-alignments-scwrl.under InfilePrefix 1v6zA/ include read-alignments-scwrl.under InfilePrefix 1vhyA/ include read-alignments-scwrl.under InfilePrefix 1q9jA/ include read-alignments-scwrl.under InfilePrefix 1zl3A/ include read-alignments-scwrl.under InfilePrefix 1odoA/ include read-alignments-scwrl.under with inc wt on break and soft clases. try8 is running on shaw I also started try9 which polishing run on T0378.try4-opt1.pdb with try5's costfcn try9 is running on vashon recap : try1-> 1x7oA, 1zjrA. try2-> try1 (polishing run) -> 1x7oA, 1zjrA. try3-> try1 and try2 (based on read-pdb.under) try4-> try1 and try2 and try3 (based on read-pdb.under) try5-> it had mistake in .under file try6-> based on try5 try7-> T0378.try4-opt2.gromacs0.repack-nonPC.pdb.gz Thu Aug 3 08:31:21 PDT 2006 try8 crashed ... so I am restarting it. try9 is scoring best now. Thu Aug 3 09:23:20 PDT 2006 Pinal Kanabar I looked at try7 carfully which was optimized from T0378.try4-opt2.gromacs0.repack-nonPC.pdb.gz as this had less breaks. Strange when I compared try7 the other models it top 1-85 residues are folded in little different manner.I had high wt on break and soft clashes in try7, so I was not sure it was due to that it just opened up N terminal. For try10 it same as try7 just with reduced break and soft clases. try10 is running on shaw Thu Aug 3 11:47:50 PDT 2006 : I messed up with try8... I forgot to replace try7 in try8.under so T0378.try7-opt2.pdb got over wrote. And now it makes sense like why try7 looks different. I will start try8 again...(try7 is same as try10 with just little change in cost function ...) Thu Aug 3 13:47:51 PDT 2006 George Shackelford I looked at the latest score-all+servers.unconstrained and try9 is at the top. Really good in breaks and soft clashes. Really good period. I don't think we need to overwork this (Pinal...); I think we can wrap this one up as well. Good work Thu Aug 3 16:15:00 PDT 2006 Pinal Kanabar Little recap : try1-> 1x7oA, 1zjrA. try2-> try1 (polishing run) -> 1x7oA, 1zjrA. try3-> try1 and try2 (based on read-pdb.under) try4-> try1 and try2 and try3 (based on read-pdb.under) try5-> it had mistake in .under file try6-> based on try5 try7-> T0378.try4-opt2.gromacs0.repack-nonPC.pdb.gz (Got overwrite with try8.under and try7.costfcn) try8-> 1gz0A, 1v2xA, 1ipaA,1mxiA,1z85A,1v6zA, 1vhyA,1q9jA,1zl3A,1odoA try9-> T0378.try4-opt2.pdb -> try1 , try2 and try3 (based read-pdb.under) try10-> T0378.try4-opt2.gromacs0.repack-nonPC.pdb Looking at score-all+servers.unconstrained.pretty try9 is doing best followed by T0378.try9-opt2.pdb.gz T0378.try3-opt2.pdb T0378.try9-opt1.pdb.gz T0378.try4-opt2.pdb.gz T0378.try4-opt1.pdb.gz T0378.try3-opt1.pdb T0378.try3-opt2.pdb.gz T0378.try3-opt1.pdb.gz T0378.try10-opt2.pdb.gz T0378.try9-opt2.repack--nonPC.pdb.gz T0378.try3-opt2.repack--nonPC.pdb.gz T0378.try4-opt2.repack-nonPC.pdb.gz T0378.try7-opt2.pdb.gz T0378.try2-opt2.pdb.gz T0378.try10-opt2.pdb.gz try8 which was made from different alignments scored really bad by undertaker but it Rosetta faovrite. My personal choice : try9, try3, try4, try10, and try8 try8 has different fold.... but still as we get to submitt five models we should submit try8.. Anyways I leave to George for final submission Thu Aug 3 16:34:48 PDT 2006 George Shackelford Pinal's choices look good. Try9 is best, try3 and try4 are worthy includes. Since try10 comes from try4 except that it does not do as well, I think I'll slip the t06 server's #1 in it's place. Oddly try8 does the best in Rosetta scoring with try7(!) and try10 as well. I'm still going to replace try10. try9, try3, try4, Ok, I need to do a write-up. I'll have to do it from home. ========================== We found very strong hits on T0378, to 1gz0A (top hit) ,1x7oA,1ipaA and others. These were from SCOP domains c.116.1.1,d.79.3.3. BLAST also got strong PDB hits; the best three were 1ipaA, 1x7oA, and 1zjrA. We judged it to be comparative modeling. Try1 was generated in the standard by using alignments from SAM-t2k, SAM-t04, and SAM-t06. It apparently was based on 1zjrA (after first preferring 1x7oA). Model 1 is try9-opt2, our best scoring model, the result of three rounds of polishing the automatic with undertaker. Model 2 is try8-opt2.gromacs0.repack-nonPC. It is the best Rosetta score; its side chains were repacked by Rosetta. Try8 was started with a set of specific alignments - HMM favorites and some of the top PDB hits: 1gz0A, 1v2xA, 1ipaA,1mxiA,1z85A,1v6zA, 1vhyA,1q9jA,1zl3A,1odoA. 1x7oA and 1ipaA were not included so as to force a possible different model. Model 3 is try3-opt2 which is based on try1,the automatic model, and try2 a polishing of try1. It try8-opt2.gromacs0.repack-nonPC.pdb Model 4 is try4-opt2. This was optimized from try1, try2, and try3. However it did not do as well as try3 in scoring; we do not know why. Model 5 is try10-opt2.gromacs0.repack-nonPC, which had the third best Rosetta score. It is based on T0378.try4-opt2.gromacs0.repack-nonPC. Both of these models used gromacs for energy minimizing and had their side chains repacked by Rosetta. Thu Aug 3 21:06:15 PDT 2006 Kevin Karplus It looks like George had set everything up , but not done the submission. Since Grant seems to be busy with T0376, I did the "make casp_models" and "make email" for T0378. Thu Aug 3 22:17:05 PDT 2006 Grant Thiltgen Thanks Kevin, I was just getting ready to take care of this. Thu Nov 2 11:05:27 PST 2006 Kevin Karplus We seem to have messed this one up. Our best result was for align2. We may have done a bit better if we evaluate the two domains separately, as we seem to have done ok on the C-terminal domain.