Tue Jun 3 09:01:59 PDT 2008 T0429 Make started Tue Jun 3 09:02:15 PDT 2008 Running on cheep.cse.ucsc.edu Tue Jun 3 09:03:08 PDT 2008 Kevin Karplus reported by crystallographer as monomer. pdb_blast has probable hit to 2nq3A. Thre is a HIS tag on the N terminus. Tue Jun 3 09:49:25 PDT 2008 Kevin Karplus The t06.w0.5 HMM also finds 2nq3A (b.7.1.1) as a good hit and nothing else. Tue Jun 3 11:01:59 PDT 2008 Kevin Karplus Other than t06.w0.5 and t04.w0.5, there doesn't seem to be much agreement among the HMMs about what the template(s) should be. THe E-values are generally quite high. This may be a "more luck than judgement" modeling target! Tue Jun 3 11:28:09 PDT 2008 Kevin Karplus The best E-value was 9.5, so the HMMs failed to find a convincing template. The best-scoring one was not 2nq3A (b.7.1.1, C2-domain-like) but 2qqrA, which is not in SCOP. Second best was 2hqxA (b.34.9.1). Several of the b.34 (SH3-like barrel) templates scored in the top few models, so this is encouraging. Wed Jun 4 09:39:09 PDT 2008 Kevin Karplus The automatic try1 model looks awful, but the first alignment, to 2qqrA looks pretty good. It seems to have two barrel domains, with one inserted into the middle of the other's sequence. The other hits seem to be for individual barrels. The evidence for the inserted domain looks pretty solid to me. We might be able to model the inner barrel (roughly R54-R143) independently, but the outer-barrel seems harder---perhaps I could make alignments by having R54-R143 x-ed out. I've not done much playing with split domains---this might be a good chance to practice. Wed Jun 4 09:48:49 PDT 2008 Kevin Karplus R54-R143 prediction started on moai cluster. Outer/ subdirectory created, and residues between R54 and R153 were replaced by lower-case x in T0429.a2m I also edited the T0429.blank.pdb.gz file and try1.under to use ReadTargetPDB T0429.blank.pdb instead of ReadTargetSeq T0429.a2m 1 Let's see whether this manages to do any modeling, or if the lower-case x letters mess everything up. Wed Jun 4 11:31:52 PDT 2008 Kevin Karplus The Outer/ run doesn't get good evalues, but does get b.34.9.1 (SH3-like barrels) as the first hit (1mhnA at E-value 12). I've also started and Outer-swapped/ prediction that swaps the two parts of the outer domain, looking to see if a permutation of the barrel fits better. The R54-R143 prediction gets 1mhnA as its second prediction (E-value 40), but it's first prediction looks wrong (d.129.3.2, evalue 30). Wed Jun 4 11:44:47 PDT 2008 Kevin Karplus I grabbed the top b.34 hits in R54-R143 and set them up as MANUAL_TOP_HITS, and started R54-R143/try2 using just those alignments. Wed Jun 4 17:55:08 PDT 2008 Kevin Karplus R54-R143/try1 and try2 are somewhat different versions of the barrel. I don't know which is better. Neither sticks out R45 and R143 as a linker to the outer domain. There are plenty of b.34 domains in the top scores, but the best E-value for any of them is 39.6. The Outer/ domain gets 1mnhA as its top hit (with 2hqxA second, both b.34.9.1), but the try1 run failed---I'll have to figure out what the discontinuous domain broke. The Outer-swapped/ domain 2imgA (not in SCOP) as its top domain, with E-value 32. There are no b.34 in the list of top hits, so I thin I'll regard Outer-swapped/ as a deadend for now. Sun Jun 8 11:06:32 PDT 2008 SAM-T08-MQAO hand QA T0429 Submitted Sun Jun 8 11:06:32 PDT 2008 SAM-T08-MQAU hand QA T0429 Submitted Sun Jun 8 11:06:32 PDT 2008 SAM-T08-MQAC hand QA T0429 Submitted Mon Jun 9 12:29:13 PDT 2008 Kevin Karplus The metaserver run for MQAU1-opt3 is based mainly on BAKER-ROBETTA_TS1 The metaserver run for MQAC1-opt3 is based mainly on BAKER-ROBETTA_TS1 The metaservers are predicting GDT values around 26-29% costfcn favorite models try1 MQAC1-opt3 MQAU1-opt3 try1-opt3.gromacs0 I need to start putting the domain predictions together to get a decent model. The BAKER-ROBETTA model does seem to be a 2qqrA-based model. Maybe I should make an extra effort using just that template. The BAKER-ROBETTA model seems to agree well with the 2qqrA alignment, and fairly well with R54-R143/try2-opt3 There is a big loop region that is variously defined, but the later parts of the barrel agree. Sat Jun 21 11:28:40 PDT 2008 Kevin Karplus Getting back to this taarget after a couple of weeks, I see that the "Outer" subdomain predictions never really completed. It looks like the alignments weren't made, so I'll take all the best hits and use them as MANUAL_TOP_HITS in Outer/Makefile and run "make extra_alignments" and "make read_alignments" in the Outer/ subdirectory. Then I'll try again to make the Outer/T0419.undertaker-align.pdb.gz file and the Outer/try1 run. Sat Jun 21 13:29:40 PDT 2008 Kevin Karplus I didn't seem to be having any trouble with making individual files in Outer/ so I'll try rerunning the whole make---there may have been intermittent troubles with file server or with a moai cluster node. Sat Jun 21 14:04:39 PDT 2008 Kevin Karplus The undertaker script to optimize constraints from alignments fails in Outer/ which means that the the try1 optimization there is not going to use alignment constraints. Sat Jun 21 14:26:39 PDT 2008 Kevin Karplus I'm trying try2 using alignments from the whole chain and the two subdomains. The costfcn is a bit of a gamble, as I'm not sure which of the competing alignmenst of the sheets to favor. Sat Jun 21 14:38:43 PDT 2008 Kevin Karplus Because of a typo in try2.under the first attempt at try2 failed. Trying again. Sat Jun 21 14:46:51 PDT 2008 Kevin Karplus I also just noticed that the Outer/try1 run did not include a KnownBreak command for the middle domain, so I added it to costfcn-init.under and am running Outer/try2 from alignments. Sat Jun 21 17:16:55 PDT 2008 Kevin Karplus Neither of the "Outer" runs looks very convincing, but the try2 run from this directory produces a decent outer domain, though the middle domain is torn apart a bit. Perhaps I should take M1-R54 Y149-E178 from try2-opt3, and K55-R143 from R54-R143/try1 or R54-R143/try2 (T144-L148 also from try2-opt3). I'll make 2 such chimeras, then optimize each with the corresponding sheet constraints. Sat Jun 21 21:07:01 PDT 2008 Kevin Karplus try3 kept the two barrels intact, but did not manage to close the gaps in the linkers between the barrels. Perhaps I need to make a new model for the central domain, with a strong constraint to hold residues together at the end: Constraint R54.CA Y149.CA -999 7.4 8 Constraint R54.NH2 Y149.OH 0 3.9 5 Hbond T53.N Y149.O Unfortuantely, the central domain that I modeled only extended to R54-R143, so does not include these residues. Maybe I should model a bigger range, so that I can include them and make a compatible, plug-in model for the central domain. Sun Jun 22 09:27:00 PDT 2008 Kevin Karplus R51-N153/try1 looks ok, but I'll try polishing it to close gaps and improve the ends before atempting to paste it into the whole model. Sun Jun 22 12:31:58 PDT 2008 Kevin Karplus I'm still not sure how to put the pieces together to make a good 2-domain model. Sun Jun 22 13:22:30 PDT 2008 Kevin Karplus I made chimera-MQAC1-M2 from MQAC1-try3 and D74-T142 of R51-N153/try2-opt3. I'll optimize it in try5 Sun Jun 22 13:58:37 PDT 2008 Kevin Karplus for chimera-try2-R51-try2 I'll take M1-D60,Y149-end from try2-opt3 and E61-L148 from R51-N151-try2-opt3 (I just noticed that try5.costfcn had the wrong sheet constraints for the outer part---I may need to redo it with MQAC1 sheets.) Sun Jun 22 17:00:41 PDT 2008 Kevin Karplus try6, based on chimera-try2-R1-try2 never got very good---not even up to the scores of try1 or try2. try7, on the other hand, looks pretty good, and offers a slightly different solution for where the N-terminus goes, allowing the loop near P125 to fold in better. I'll do one more polishing run from MQA* (try8) and one from try7 and try5 (try9). Those will have to be my top 2 choices. I wish I had more time to explore other psosibilities. costfcn top models try9 try8-opt3 MQAU1-opt3.gromacs0 MQAC1-opt3.gromacs0 try9-opt3 try7-opt3.gromacs0 try8 try8-opt3 MQAU1-opt3.gromacs0 MQAC1-opt3.gromacs0 try9-opt3 try7-opt3.gromacs0 try1 try8-opt3 MQAC1-opt3 MQAU1-opt3 try9-opt3 try1-opt3.gromacs0 try7-opt3.gromacs0 I'll submit ReadConformPDB T0429.try8-opt3.pdb < MQAC1-opt3 < BAKER-ROBETTA_TS1 ReadConformPDB T0429.try9-opt3.pdb < try7-opt3 < chimera-MQAC1-M2 < outer part from T0429.MQAC1-opt3.pdb < D74-T142 from R51-N151/try2-opt3 ReadConformPDB T0429.MQAU1-opt3.gromacs0.pdb < BAKER-ROBETTA_TS1 ReadConformPDB T0429.MQAC1-opt3.gromacs0.pdb < BAKER-ROBETTA_TS1 ReadConformPDB T0429.try1-opt3.gromacs0.pdb < aligments (various used) Tue Jul 8 14:25:55 PDT 2008 Kevin Karplus T0429 was chosen as a refinement target. The notes on that are in ../TR429 The GDT of "one of the best models" was only 47%, still better than the predicted GDT from the servers of 26-30%. I'll try finding out which of our models or the server models comes closest: make everything.costfcn ; undertaker < find-TR429.under >& find-TR429.log; gzip -9f find-TR429.log The closest models to TR429 are by GDT: RAPTOR_TS3 GS-KudlatyPred_TS5 Phragment_TS5 try1-opt3.gromacs0.pdb COMA-M_TS1 by real_cost: Phragment_TS5 RAPTOR_TS3 try1-opt3.gromacs0.repack-nonPC Phyre_de_novo_TS1 It looks like we did not do so well on this target (though we did submit a try1 model as our model 5, which may not be too terrible).