Tue May 30 10:42:52 PDT 2006 T0305 Make started Tue May 30 10:44:39 PDT 2006 Running on orcas.cse.ucsc.edu Tue May 30 11:06:51 PDT 2006 Kevin Karplus Excellent blast hit to 2fh7A (E-value 6.6e-80). Actually hits 2fh7A twice---the template seems to have 2 copies of the domain. There are 62 PDB matches in the t06 multiple alignment, several of which occur twice (for two-domain templates). Tue May 30 19:22:07 PDT 2006 Kevin Karplus Top HMM hits are 2fh7A, 1pa1A, 2f71A, 2b49A, 2bzlA, 1wchA, 2bv5A, ... Thu Jun 15 11:48:01 PDT 2006 Kevin Karplus Top scores with unconstrained costfcn are T0305.try1-opt2.pdb.gz SAM_T06_server_TS1 ROBETTA_TS4 ROBETTA_TS3 ROBETTA_TS1 ROBETTA_TS5 ROBETTA_TS2 PROTINFO_TS1 PROTINFO_TS2 I should do a polishing run from our models and a polishing run from the ROBETTA models. Thu Jun 15 12:01:42 PDT 2006 Kevin Karplus try2 started on cheep with our models as starting points try3 started on orcas with ROBETTA, PROTINFO, and GeneSilicoMetaServer models as starting points. Thu Jun 15 17:47:15 PDT 2006 Kevin Karplus try2-opt2 (based on our models) scores slightly better than try3-opt2, which is based on ROBETTA_TS4, but robetta prefers repacking try3-opt2. I created a try4.costfcn that scores try3-opt2 and try2-opt2 almost identically, and am optimizing (with crossover) from all our try* models. (try4 started on lopez) Fri Jun 16 07:52:16 PDT 2006 Kevin Karplus try4 seems to be based mainly on try3. I'm pretty sure the differences are now smaller than my costfcns are capable of distinguishing, so I will submit try4-opt2 try2-opt2 try3-opt2.repack-nonPC align1 from 2fh7A align2 from 1pa1A Fri Jun 16 08:04:17 PDT 2006 Kevin Karplus So submitted. Sat Jun 17 10:47:44 PDT 2006 Kevin Karplus Last night, they told us that they are accepting multimeric models for T0305, so now I need to come up with a multimer. Interestingly, the 2fh7A "dimer" is actually a single chain with two copies of the domain. There is a a bit of difference between the two C-terminal parts of the domains. It will be an interesting starting point. The second highest hit (1pa1A) is monomeric, according to both PDB and PQS, as is the third hit (2f71A). There is no obvious hydrophobic patch on the surface, so I'm not so sure I want to try to make a multimer of this one. Sat Jun 17 11:16:55 PDT 2006 Kevin Karplus I have to figure out how to get an acceptable alignment to the second domain of 2fh7A. Hmm, it looks like blast got two alignments to 2fh7A, so I could hand-edit those. >2fh7A 595 XRAY 2.00 0.192 0.238 Receptor-type tyrosine-protein phosphatase S [HOMO SAPIENS] Length = 595 Score = 292 bits (747), Expect = 7e-80 Identities = 149/286 (52%), Positives = 187/286 (65%), Gaps = 16/286 (5%) Query: 10 KHIGELYSNNQHGFSEDFEEVQRCTADMNITAEHSNHPENKHKNRYINILAYDHSRVKLR 69 +H L +N+ S+++E + T EHSN NK KNRY N++AYDHSRV L+ Sbjct: 19 EHTERLKANDSLKLSQEYESID---PGQQFTWEHSNLEVNKPKNRYANVIAYDHSRVILQ 75 Query: 70 PLPGKDSKHSDYINANYVDGYNKAKAYIATQGPLKSTFEDFWRMIWEQNTGIIVMITNLV 129 P+ G SDYINANYVDGY + AYIATQGPL TF DFWRM+WEQ + IVM+T L Sbjct: 76 PIEG--IMGSDYINANYVDGYRRQNAYIATQGPLPETFGDFWRMVWEQRSATIVMMTRLE 133 Query: 130 EKGRRKCDQYWPTENSEEYGNIIVTLKSTKIHACYTVRRFSIRNTKVKKGQKGNPKGRQN 189 EK R KCDQYWP +E YG I VTL T A + VR FS+ G Sbjct: 134 EKSRIKCDQYWPNRGTETYGFIQVTLLDTIELATFCVRTFSLHKN-----------GSSE 182 Query: 190 ERVVIQYHYTQWPDMGVPEYALPVLTFVRRSSAARMPETGPVLVHCSAGVGRTGTYIVID 249 +R V Q+ +T WPD GVPEY P L F+RR P+ GP++VHCSAGVGRTG +IVID Sbjct: 183 KREVRQFQFTAWPDHGVPEYPTPFLAFLRRVKTCNPPDAGPIVVHCSAGVGRTGCFIVID 242 Query: 250 SMLQQIKDKSTVNVLGFLKHIRTQRNYLVQTEEQYIFIHDALLEAI 295 +ML++IK + TV+V G + +R+QRNY+VQTE+QY FIH+ALLEA+ Sbjct: 243 AMLERIKPEKTVDVYGHVTLMRSQRNYMVQTEDQYSFIHEALLEAV 288 Score = 211 bits (537), Expect = 1e-55 Identities = 108/252 (42%), Positives = 157/252 (62%), Gaps = 15/252 (5%) Query: 44 SNHPENKHKNRYINILAYDHSRVKLRPLPGKDSKHSDYINANYVDGYNKAKAYIATQGPL 103 +N P NK KNR +NI+ Y+ +RV L+P+ G + SDYINA+++DGY + KAYIATQGPL Sbjct: 339 ANLPCNKFKNRLVNIMPYESTRVCLQPIRGVEG--SDYINASFIDGYRQQKAYIATQGPL 396 Query: 104 KSTFEDFWRMIWEQNTGIIVMITNLVEKGRRKCDQYWPTENSEEYGNIIVTLKSTKIHAC 163 T EDFWRM+WE N+ I+VM+T L E GR KC QYWP E S Y +V A Sbjct: 397 AETTEDFWRMLWENNSTIVVMLTKLREMGREKCHQYWPAERSARYQYFVVDPM-----AE 451 Query: 164 YTVRRFSIRNTKVKKGQKGNPKGRQNERVVIQYHYTQWPDMGVPEYALPVLTFVRR--SS 221 Y + ++ +R KV + G R V Q+ +T WP+ GVP+ + F+ + + Sbjct: 452 YNMPQYILREFKVTDARDG------QSRTVRQFQFTDWPEQGVPKSGEGFIDFIGQVHKT 505 Query: 222 AARMPETGPVLVHCSAGVGRTGTYIVIDSMLQQIKDKSTVNVLGFLKHIRTQRNYLVQTE 281 + + GP+ VHCSAGVGRTG +I + +L++++ + V++ +K +RTQR +VQTE Sbjct: 506 KEQFGQDGPISVHCSAGVGRTGVFITLSIVLERMRYEGVVDIFQTVKMLRTQRPAMVQTE 565 Query: 282 EQYIFIHDALLE 293 ++Y F + A LE Sbjct: 566 DEYQFCYQAALE 577 I could also try 1larA: >1larA 575 XRAY 2.00 0.222 0.274 LAR [HOMO SAPIENS] || 1LARB Length = 575 Score = 280 bits (717), Expect = 2e-76 Identities = 143/284 (50%), Positives = 186/284 (65%), Gaps = 16/284 (5%) Query: 11 HIGELYSNNQHGFSEDFEEVQRCTADMNITAEHSNHPENKHKNRYINILAYDHSRVKLRP 70 +I L +N+ FS+++E + T E+SN NK KNRY N++AYDHSRV L Sbjct: 8 NIERLKANDGLKFSQEYESID---PGQQFTWENSNLEVNKPKNRYANVIAYDHSRVILTS 64 Query: 71 LPGKDSKHSDYINANYVDGYNKAKAYIATQGPLKSTFEDFWRMIWEQNTGIIVMITNLVE 130 + G SDYINANY+DGY K AYIATQGPL T DFWRM+WEQ T +VM+T L E Sbjct: 65 IDGVPG--SDYINANYIDGYRKQNAYIATQGPLPETMGDFWRMVWEQRTATVVMMTRLEE 122 Query: 131 KGRRKCDQYWPTENSEEYGNIIVTLKSTKIHACYTVRRFSIRNTKVKKGQKGNPKGRQNE 190 K R KCDQYWP +E G I VTL T A YTVR F++ + G + Sbjct: 123 KSRVKCDQYWPARGTETCGLIQVTLLDTVELATYTVRTFALHKS-----------GSSEK 171 Query: 191 RVVIQYHYTQWPDMGVPEYALPVLTFVRRSSAARMPETGPVLVHCSAGVGRTGTYIVIDS 250 R + Q+ + WPD GVPEY P+L F+RR A + GP++VHCSAGVGRTG +IVID+ Sbjct: 172 RELRQFQFMAWPDHGVPEYPTPILAFLRRVKACNPLDAGPMVVHCSAGVGRTGCFIVIDA 231 Query: 251 MLQQIKDKSTVNVLGFLKHIRTQRNYLVQTEEQYIFIHDALLEA 294 ML+++K + TV++ G + +R+QRNY+VQTE+QY+FIH+ALLEA Sbjct: 232 MLERMKHEKTVDIYGHVTCMRSQRNYMVQTEDQYVFIHEALLEA 275 Score = 206 bits (523), Expect = 6e-54 Identities = 107/252 (42%), Positives = 156/252 (61%), Gaps = 15/252 (5%) Query: 44 SNHPENKHKNRYINILAYDHSRVKLRPLPGKDSKHSDYINANYVDGYNKAKAYIATQGPL 103 +N P NK KNR +NI+ Y+ +RV L+P+ G + SDYINA+++DGY + KAYIATQGPL Sbjct: 327 ANLPCNKFKNRLVNIMPYELTRVCLQPIRGVEG--SDYINASFLDGYRQQKAYIATQGPL 384 Query: 104 KSTFEDFWRMIWEQNTGIIVMITNLVEKGRRKCDQYWPTENSEEYGNIIVTLKSTKIHAC 163 + EDFWRM+WE N+ IIVM+T L E GR KC QYWP E S Y +V A Sbjct: 385 AESTEDFWRMLWEHNSTIIVMLTKLREMGREKCHQYWPAERSARYQYFVVDPM-----AE 439 Query: 164 YTVRRFSIRNTKVKKGQKGNPKGRQNERVVIQYHYTQWPDMGVPEYALPVLTFVRR--SS 221 Y + ++ +R KV + G R + Q+ +T WP+ GVP+ + F+ + + Sbjct: 440 YNMPQYILREFKVTDARDG------QSRTIRQFQFTDWPEQGVPKTGEGFIDFIGQVHKT 493 Query: 222 AARMPETGPVLVHCSAGVGRTGTYIVIDSMLQQIKDKSTVNVLGFLKHIRTQRNYLVQTE 281 + + GP+ VHCSAGVGRTG +I + +L++++ + V++ +K +RTQR +VQTE Sbjct: 494 KEQFGQDGPITVHCSAGVGRTGVFITLSIVLERMRYEGVVDMFQTVKTLRTQRPAMVQTE 553 Query: 282 EQYIFIHDALLE 293 +QY + A LE Sbjct: 554 DQYQLCYRAALE 565 but for make-dimer1, I'll just use the HMM-based alignment to the first domain of 2fh7A and the BLAST-based alignment to the second domean of 2fh7A. I hand edited the sequence to get the BLAST alignment as the second alignment in 2fh7A/T0305-2fh7A.dimer.a2m Sat Jun 17 11:55:10 PDT 2006 Kevin Karplus Well, I made the "dimer" based on 2fh7A, and it doesn't look like a dimer to me! It looks like half a tetramer. I think I'll use the 4 domains of 1larA and 1larB as the template for making the tetramer. Sat Jun 17 12:17:29 PDT 2006 Kevin Karplus I created 1larA/T0305-1larA.tetramer.a2m and make-tetramer1.under, then ran undertaker < make-tetramer1.under to create tetramer/decoys/tetramer-try4-1larA.pdb.gz I created tetramer/T0305.a2m, tetramer/Makefile, and tetramer/try1.under (edited from the standard try1.under to read the alignments and Template.atoms from ../, having TweakMultimer turned up, starting from the one tetramer we made). I copied try4.costfcn to tetramer/try1.costfcn, and removed the references to the predicted alpha values (using just the propensities alpha and alpha_prev). Sat Jun 17 12:54:54 PDT 2006 Kevin Karplus Oops, I forgot to include the "multimer" keyword in the Optconform for tetramer/try1.under. There may be some advantage to this mistake, as I'm not sure that the tetramer is really from the symmetry group S4. That is, the transformation chain A->B may not be 4th root of the identity transform, though A->C is a square root. I'll start a try2 that is identical with try1, except that it will try to make 4-fold symmetry, and try3 that tries to make 2-fold symmetry. Sat Jun 17 13:01:54 PDT 2006 Kevin Karplus try2 and try3 started on lopez. Sat Jun 17 15:22:13 PDT 2006 Kevin Karplus tetramer/try1-opt1 has changed to a cyclic tetramer, even without instructions in the Optconform---I think that the marking of the initial model as a tetramer was enough to cause the symmetrizing. Sat Jun 17 18:45:41 PDT 2006 Kevin Karplus The tetramer is too large for Rosetta to repack: STOP:: don MAX_HB_ATM exceeded Sat Jun 17 20:16:27 PDT 2006 Kevin Karplus All the attempts to improve the tetramer as a tetramer resulted in a cyclic 4-mer. While we may want to submit one such model, none of try1, try2, or try3 are very convincing. I'll try again, making an additional effort to treat the model as a big dimer. (There was still one place in try3 where I'd left in a "multimer 4") tetramer/try4 started on whidbey Sat Jun 17 21:17:39 PDT 2006 Kevin Karplus In tetramer/try4.log, I see that the big gains (loss of clashes) seem to come first from OptSubtree, then from TweakMultimer. This leads me to suspect that an OptSubtree is changing the A->B transform, and TweakMultimer is changing the AB->CD transform, which is what I was hoping for. I'll not be able to *see* it until the tetramer/try4-opt1 conformation is output, but I'm hopeful ... . Sun Jun 18 06:31:20 PDT 2006 Kevin Karplus tetramer/try4 has indeed modified the A->B relationship and the AB->CD relationship separately. Unfortunately, it has still moved things too far apart, so that the tetramer is not tightly packed. I don't know *how* I'm going to get a tightly packed tetramer, as every measure (even phobic fit) prefers the various tries over the initial tetramer. Sun Jun 18 13:17:45 PDT 2006 Kevin Karplus Are there other choices for making a tetramer? We could try making a dimer from 1rpm[AB]. It is really just 2 monomors packed into the crystal. Sun Jun 18 13:40:48 PDT 2006 Kevin Karplus I made a dimer (using make-dimer2.under) from 1rpmA and have started optimizing it (and the older 2fh7A dimer) in dimer/try1. I'm trying something a little different there---treating the dimer from 2fh7A as a monomer, and the one from 1rpmA as a dimer. If undertaker is working correctly, the derived models should keep track of whether they need to symmetrize or not. Sun Jun 18 13:48:31 PDT 2006 Kevin Karplus dimer/try1 seems to be working mainly on 2fh7A, so I might want to start a dimer/try2 that works only on 1rpmA. (started dimer/try2 on orcas) Sun Jun 18 17:06:49 PDT 2006 Kevin Karplus dimer/try1 finished, but try2-opt1 already scores better than try2-opt2, mainly due to differences in phobic_fit Sun Jun 18 20:18:40 PDT 2006 Kevin Karplus I think that for oligomers I want to submit: monomer try4-opt2 tetramer tetramer-try4-1larA.pdb dimer try2-opt2 dimer try1-opt2 tetramer try1-opt2 But I still want to find out whether monomers can be submitted as oligomeric predictions. Sun Jun 18 21:05:09 PDT 2006 Kevin Karplus Actualy, we have 3 distinct dimer solutions: try2-opt2 try1-opt2 dimer-try4-1rpmA I may want to eliminate the monomer solution in favor of 3 dimers and 2 tetramers. Mon Jun 19 11:18:45 PDT 2006 Kevin Karplus Only dimeric unit cells will be accepted for CASP for this target, so I need to reduce the tetramers to dimeric unit cells. I think I'll take tetramer/ tetramer-try4-1larA.pdb and tetramer/ try1-opt2 chains A & B and put them in the dimer directory. I'll then optimize them as non-cyclic dimers Mon Jun 19 11:50:39 PDT 2006 Kevin Karplus Currently, I favor the following 5 dimers, though from-tetramer-try4-1larA.pdb and T0305.try1-opt2.pdb are very similar. ReadConformPDB T0305.try1-opt2.pdb ReadConformPDB T0305.try2-opt2.pdb ReadConformPDB from-tetramer-try1-opt2.pdb ReadConformPDB dimer-try4-1rpmA.pdb ReadConformPDB from-tetramer-try4-1larA.pdb I may want to try polishing from-tetramer-try1-opt2.pdb Nah, it's not worth bothering with---the small changes we are likely to get are not going to worth it. I'll just submit the current 5 dimers. Mon Jun 19 12:17:32 PDT 2006 Kevin Karplus So submitted. Fri Jul 14 11:54:33 PDT 2006 Kevin Karplus With evaluate.unconstrained.pretty, Best server: FOLDpro_TS1 -1.60 Our best model; try4-opt2.gromacs0 -1.46 Our best submitted: model1 -1.46 Our best server:SAM_T06_server -1.01 32 of 51 TS1