Mon Jun 23 09:30:04 PDT 2008 T0467 Make started Tue Jun 24 09:25:08 PDT 2008 Running on moai01.kilokluster.ucsc.edu Tue Jun 24 11:21:04 PDT 2008 Kevin Karplus template-free? best e-value is 2.4. b.34.13.1 looks like a possible hit. Note: this has many of the same hits as T0468, so these two proteins should be worked on together. Tue Jun 24 17:26:36 PDT 2008 Kevin Karplus try1-opt3 looks like a good beginning (better than T0468 managed), so I'll pass it on to VAST to see what templates it may be based on. Request id: 99650805835832867 http://www.ncbi.nlm.nih.gov/Structure/vast/VSMmdb.cgi?cmdVSMmdb=StrText&grpid=99650805835832867&ViewNbr=Yes VAST reports so-so hits to 9 pdb chains in the medium-redundancy set: 1l0oA 2oggA 1wgvA 2cr0A 1qpkA 2joeA 1q16B 2aihA 3bxoA (1q16B=1y5iB, the rest are in dunbrack-pdbaa) None of these came up as hits in the HMMs, so I 'm not sure it is worth trying to force alignments to them to try to get different templates. Maybe ... Tue Jul 1 15:57:24 PDT 2008 SAM-T08-MQAO hand QA T0467 Submitted Tue Jul 1 15:57:24 PDT 2008 SAM-T08-MQAU hand QA T0467 Submitted Tue Jul 1 15:57:24 PDT 2008 SAM-T08-MQAC hand QA T0467 Submitted Mon Jul 21 11:55:17 PDT 2008 Kevin Karplus I haven't even done the meta-server runs yet for this target! GULP! I'm way behind schedule! Mon Jul 21 13:30:22 PDT 2008 Kevin Karplus MQAC1 and MQAU1 both based on SAM-T08-server_TS1 I need to do an MQAY1 that excludes the SAM servers. Mon Jul 21 13:38:57 PDT 2008 Kevin Karplus The C-terminus of MQAU1, MQAC1, and try1 agrees with the top alignments. But the N-terminus (up to K38) is not in the top alignments. If the MQAY run comes up with similar models, then I *should* take the best of them and send it off to VAST to get some longer templates. Mon Jul 21 13:58:58 PDT 2008 Kevin Karplus MQAY1 is optimizing BAKER-ROBETTA_TS5 Mon Jul 21 17:48:10 PDT 2008 Kevin Karplus The MQAY1 model fits the secondary structure prediction better than the other models. I wonder if I could use the first 38 or so residues of that with MQAC1-opt3 to get a better combined model. I'd have to do a different superposition than the best-models one. Mon Jul 21 22:07:38 PDT 2008 Kevin Karplus I tried playing around with an MQAY1-MQAC1 chimera, but I couldn't get the pieces to go together. I looked at BAKER-ROBETTA_TS1, and it looks a lot like the model I'm now favoring for T0468, so I'll grab constraints from it and do a metaserver run. Mon Jul 21 22:14:33 PDT 2008 Kevin Karplus MQAY2 run started on moai cluster. Tue Jul 22 11:49:54 PDT 2008 Kevin Karplus As expected, the MQAY2 run polishes BAKER-ROBETTA_TS1, improving the packing and Hbonds a little bit. I've started a try2 run to continue the polishing with the same costfcn (mainly so that I don't have a gap in the numbering---I don't really expect anything to move much). I'll submit the MQAY2 model to VAST, and add the hits to my MANUAL_TOP_HITS list, then try doing a run from alignments. Your VAST Search job was submitted at 07/22/2008 15:03:38(EDT). Request ID: 849425583072305863 Tue Jul 22 12:18:41 PDT 2008 Kevin Karplus Top VAST hits by aligned length: PDB C D Ali. Len. SCORE P-VAL RMSD %Id Description 1PRN A 79 8.9 0.0198 3.0 3.8 Refined Structure Of Porin From Rhodopseudomonas Blastica And Comparison With The Porin From Rhodobacter Capsulatus 2D1E A 1 74 8.3 0.0015 3.0 9.5 Crystal Structure Of Pcya-Biliverdin Complex 2D1E A 73 8.3 0.0047 2.8 9.6 Crystal Structure Of Pcya-Biliverdin Complex 2POR A 73 9.7 0.0051 2.9 5.5 Structure Of Porin Refined At 1.8 Angstroms Resolution 1DCU A 1 62 9.0 10e-4.9 2.8 1.6 Redox Signaling In The Chloroplast: Structure Of Oxidized Pea Fructose-1,6-Bisphosphate Phosphatase 1DCU A 61 9.0 0.0387 2.7 1.6 Redox Signaling In The Chloroplast: Structure Of Oxidized Pea Fructose-1,6-Bisphosphate Phosphatase 2O62 A 2 57 8.1 0.0183 2.9 7.0 Crystal Structure Of Hypothetical Protein (Zp_00105914.2) From Nostoc Punctiforme Pcc 73102 At 1.75 A Resolution 2UV8 G 8 56 7.6 0.0400 2.7 5.4 Crystal Structure Of Yeast Fatty Acid Synthase With Stalled Acyl Carrier Protein At 3.1 Angstrom Resolution 2E0W B 1 55 9.3 0.0269 2.0 0.0 T391a Precursor Mutant Protein Of Gamma- Glutamyltranspeptidase From Escherichia Coli 1MHM A 2 55 7.5 0.0113 1.7 3.6 Crystal Structure Of S-Adenosylmethionine Decarboxylase From Potato by Pvalue: 1Q4O B 1 52 9.4 10e-5.9 1.9 5.8 The Structure Of The Polo Box Domain Of Human Plk1 1DCU A 1 62 9.0 10e-4.9 2.8 1.6 Redox Signaling In The Chloroplast: Structure Of Oxidized Pea Fructose-1,6-Bisphosphate Phosphatase 1NFI E 2 28 8.6 10e-4.8 2.2 0.0 I-Kappa-B-AlphaNF-Kappa-B Complex 2BSZ A 1 38 7.0 10e-4.6 1.3 7.9 Structure Of Mesorhizobium Loti Arylamine N- Acetyltransferase 1 3B95 A 1 27 9.1 10e-4.5 2.3 7.4 Euhmt1 (Glp) Ankyrin Repeat Domain (Structure 2) 1NFI E 31 9.7 10e-4.3 2.7 3.2 I-Kappa-B-AlphaNF-Kappa-B Complex 1K3Z D 1 24 9.0 10e-4.2 2.3 0.0 X-Ray Crystal Structure Of The IkbbNF-Kb P65 Homodimer Complex 1N11 A 2 24 8.4 10e-4.0 2.7 4.2 D34 Region Of Human Ankyrin-R And Linker 2R19 B 42 9.9 0.0001 1.3 0.0 Crystal Structure Of The Periplasmic Lipopolysaccharide Transport Protein Lpta (Yhbn), Orthorhombic Form 2R19 B 46 9.8 0.0001 1.4 13.0 Crystal Structure Of The Periplasmic Lipopolysaccharide Transport Protein Lpta (Yhbn), Orthorhombic Form 2QC9 B 1 24 8.3 0.0001 2.4 4.2 Mouse Notch 1 Ankyrin Repeat Intracellular Domain 2QC9 B 32 9.2 0.0001 4.0 6.2 Mouse Notch 1 Ankyrin Repeat Intracellular Domain 2YZY A 2 49 8.2 0.0002 1.6 0.0 Crystal Structure Of Uncharacterized Conserved Protein From Thermus Thermophilus Hb8 1S70 B 30 9.3 0.0002 2.4 6.7 Complex Between Protein SerTHR PHOSPHATASE-1 (Delta) And The Myosin Phosphatase Targeting Subunit 1 (Mypt1) 1Q4O B 2 50 8.1 0.0002 1.6 4.0 The Structure Of The Polo Box Domain Of Human Plk1 1K1A A 31 9.2 0.0002 2.6 9.7 Crystal Structure Of The Ankyrin Repeat Domain Of Bcl-3: A Unique Member Of The Ikappab Protein Family So possible hits to include: 1dcuA 1q40B 1nfiE 2r19B=2r19A 1prnA 2dieA 2o62A 2bszA 3b95A 1n11A 2qc9B=2qc9A 1s70B Tue Jul 22 15:23:51 PDT 2008 Kevin Karplus The two best models are try2 and MQAY1, derived from different BAKER-ROBETTA models. I like how MQAY1 curls the sheets around the helix, but there is a break in the sheet. Perhaps I need to add SheetConstraint E21 G29 T40 Y32 Hbond K25 and reoptimize MQAY1. Tue Jul 22 15:32:31 PDT 2008 Kevin Karplus I started that optimization as try4, and using the same costfcn from server models as MQAY4. Tue Jul 22 19:55:44 PDT 2008 Kevin Karplus rosetta still likes MQAY2-opt3.gromacs0.repack-nonPC best. try4 likes try4-opt3, then MQAY4-opt3, then MQAY1-opt3. I think I like the MQAY4-opt3 model best myself. Tue Jul 22 20:21:16 PDT 2008 Kevin Karplus I'll also try polishing up the 1vpmA models in try5, since they are also somewhat promising. I like the curled up sheets better than very flat ones. Wed Jul 23 12:37:23 PDT 2008 Kevin Karplus try5 optimizes MQAC1-opt3, but rosetta still doesn't like it. I can't say that I like it all that much myself, but it is worth including as a conjecture, in case the meander of the robetta models is wrong. I guess my main task here is to come up with another model worth submitting. Wed Jul 23 13:22:43 PDT 2008 Kevin Karplus I could try picking up the best model from T0468 and using it as a template. I made some alignments of T0467 vs T0468 using one-track and three-track t06 models. The alignments don't quite agree, but I think I like the 3-track better. I also made two alignments in the T0468 directory using the models there, but I think I like the T0467 three-track alignment best. (They agree on all but the placement of the extra 5 residues of the T0468 model between E21 and E36.) Wed Jul 23 13:34:09 PDT 2008 Kevin Karplus For try6, I'll attempt to use just the T0468.try2-opt3 model as a template. Wed Jul 23 14:20:35 PDT 2008 Kevin Karplus I keep getting segmentation faults when running try6, with no assertion failure or error report. I'll have to run under the debugger to see I I can spot the problem. Wed Jul 23 16:53:36 PDT 2008 Kevin Karplus The problem was that there were lots of constraints, but none of them were in the costfcn, so the random constraint being returned was null, which ReduceConstraint was not set up to handle. Thas has now been fixed. Wed Jul 23 18:13:35 PDT 2008 Kevin Karplus The try6 model based on alignment to T0468 is not very convincing, having huge gaps and one of the strands pulled out of the sheet. Wed Jul 23 21:12:38 PDT 2008 Kevin Karplus K33-D71 look ok, but are they any different from other models? Yes, a little bit---I could try pasting K33-D71 into MQAY2-opt3.gromacs0.repack-nonPC, which seems to be the closest model in this region. Wed Jul 23 21:45:18 PDT 2008 Kevin Karplus try7 running to polish chimera-MQAY2-try6 One of the strands on try6 is flipped relative to MQAY2, so I had to play around with the phase of the last strand. I probalby got it wrong. Wed Jul 23 21:49:57 PDT 2008 Kevin Karplus I think I want to try pasting the middle of try6 into MQAY4-opt3, which may fit better (I may have identified the wrong model in putting together the previous chimera). Wed Jul 23 22:02:29 PDT 2008 Kevin Karplus try8 started to polish chimera-MQAY4-try6 Thu Jul 24 10:23:04 PDT 2008 Kevin Karplus With unbiased cost functions (like try6 or rosetta), try7-opt3 scores better than try8-opt3. I am now choosing among 6 models model 1 is called T0467.MQAY4-opt3.pdb model 2 is called T0467.try4-opt3.pdb model 3 is called T0467.MQAY2-opt3.gromacs0.repack-nonPC.pdb model 4 is called T0467.try7-opt3.pdb model 5 is called T0467.try8-opt3.pdb model 6 is called T0467.try5-opt3.pdb try5 is distinctly different from the rest---I'll want to keep it in as an extra option. The other five all agree closely for M1-T40, but have somewhat different flexing of the end of the hairpin for T40-E50. Models 1 and 2 (MQAY4 and try4) pair F64 with Q56. Models 4 and 5 (try7 and try8) pair F64 with T54. Model 3 pairs T54 with A63, having the strand flipped over. There are two different versions of the C-terminal helix, with models 3 and 4 having one choice and 1,2, and 5 having the other. It seems like models 1 and 2 are the most similar pair I'm considering, differing mainly in the loop G41-K49 and in the curl of the C-terminal end of the sheet. I think I like the look in try4 better than the one in MQAY4. Perhaps I should polish try7 and try8 separately, then submit. Thu Jul 24 10:49:56 PDT 2008 Kevin Karplus try9 attempting to polish try7 try10 attempting to polish try8 Thu Jul 24 12:05:59 PDT 2008 Kevin Karplus Submitted with comments: The SAM HMMs came up with only weak hits for this taret (best was 1c8cA with E-value 2.4), and those hits did not seem to lead to good full-length models, so I ended up working mainly with metaserver models. The BAKER-ROBETTA_TS1 and BAKER-ROBETTA_TS5 models turned out to be the ones I liked best and used most heavily in creating my models. Since T0467 and T0468 are quite similar sequences, I tried to find similar folds for both targets. Models 3 and 4 resulted from trying to use my current favorite T0468 model as a template for T0467. I did not get a good full-length model that way, but one part of it (K33-S70) looked good, so I tried pasting that into two of my favorite full-length models and reoptimizing. Model 1 T0467.try4-opt3.pdb # < try2-opt3.gromacs0 < MQAY2-opt3 < BAKER-ROBETTA_TS1 # this claims to be from BAKER-ROBETTA_TS1, but looks more # like the TS5 models. Perhaps there was crossover in # the try2 or try4 optimization 2 T0467.MQAY2-opt3.gromacs0.repack-nonPC.pdb # < BAKER-ROBETTA_TS1 # best rosetta energy 3 T0467.try9-opt3.pdb # < try7-opt3 < chimera-MQAY2-try6 chimera-MQAY2-try6: M1-Y32, D71-K97 from T0467.MQAY2-opt3.gromacs0.repack-nonPC.pdb K33-S70 from T0467.try6-opt3.pdb try6-opt3 < align(T0468.try2-opt3) 4 T0467.try10-opt3.pdb # < try8-opt3 < chimera-MQAY4-try6 chimera-MQAY4-try6: M1-Y32, D71-K97 from T0467.MQAY4-opt3.pdb K33-S70 from T0467.try6-opt3.pdb try6-opt3 < align(T0468.try2-opt3) 5 T0467.try5-opt3.pdb # < MQAC1-opt3 < SAM-T08-server_TS1 # this was probably the best of the SAM+undertaker models, # though it came via the SAM-T08 server.