Wed May 7 09:41:29 PDT 2008 T0391 Make started Wed May 7 09:54:35 PDT 2008 Running on shaw.cse.ucsc.edu Finally, a target that is not a trivial find for BLAST! The name "CESG, rieske ferredoxin, mouse" implies that it has a ferredoxin fold---let's see if that is what the searches predict. Wed May 7 12:37:58 PDT 2008 Kevin Karplus I'm getting the strongest hits for 1uliA, which is an ISP domain (b.33.1) and a Bet v1-like (d.129.3). It looks like the hits are to the b.33.1 fold. There are a pair of highly conserved CYS residues: C57 and C80, along with two highly conserved HIS: H59 and H83. This looks suspiciously like a zinc binding site. I should check that they cluster in the initial models, then add constraints to get the right geometry around the unrepresented ion. Hmm---the information provided by the crystallographers is that it binds iron and sulphur. No mention of zinc. Wed May 7 14:44:56 PDT 2008 Kevin Karplus Try1 made a disulfide out of C57-C80, which I should probably get rid of by turning off maybe_ssbond. Wed May 7 17:48:47 PDT 2008 Kevin Karplus The top hit, 1uliA, has the same Fe2S2 binding as we expect at the CxH CxxH site as we expect for Rieske iron-sulfur proteins: http://metallo.scripps.edu/promise/RIESKE.html We can get some strong constraints from that: C57.SG C80.SG 3.80 C57.SG H59.ND1 5.34 C57.SG H83.ND1 6.56 C80.SG H59.ND1 6.35 C80.SG H83.ND1 5.63 H59.ND1 H83.ND1 3.06 The conserved L64 and F87 are in the second shell: C57.CB L64.CG 4.3 C80.CB L64.CA 5.5 Hard to get distance, because corresponding hydrophobic is a Y in 1uliA C57.SG I87. C80.CB I87. The conserved I78 is in the third shell: I78.CB L64.CG 3.6 I78.C I87.CD 4.5 ? Wed May 7 18:29:29 PDT 2008 Kevin Karplus I wonder whether H120, R39, R118 is another binding site for something. Wed May 7 20:50:43 PDT 2008 Kevin Karplus The C and H residues seem reasonably deployed in try2-opt3, but there seems to be some damage to the surrounding structure. Thu May 8 08:49:44 PDT 2008 Kevin Karplus try3 removed the constraints on the first and second shell residues, but ended up not scoring as well (even on the try3 measure) as try2-opt3. Breaks are pretty bad in try3-opt3. Perhaps it is time for a polishing run, without long fragments. Mon May 12 14:50:20 PDT 2008 Kevin Karplus try4-opt3 looks pretty good to me. A few bad clashes and breaks still. Tue May 13 11:52:46 PDT 2008 Kevin Karplus The MQAU1-opt3 and MQAC1-opt3 score a little better than the try4-opt3 model with the try4 costfcn, but lack the Hbonds for E68-D71 that attaches it to the sheet. On the other hand, they gain Hbonds for C77-C79 attaching that strand. It may be possible to do both. Should I create a chimera that attempts it? I could splice L64-G74 out of try4-opt3 into MQAC1-try3. Tue May 13 13:54:03 PDT 2008 Kevin Karplus I could also splice Q75-S98 out of the MQA models into try4-opt3 chimeras: mostly L64-G74 chimera-MQAU1-try4 MQAU1-opt3 try4-opt3 chimera-MQAC1-try4 MQAC1-opt3 try4-opt3 mostly Q75-S98 chimera-try4-MQAU1 try4-opt3 MQAU-opt3 chimera-try4-MQAC1 try4-opt3 MQAC-opt3 Tue May 13 14:13:01 PDT 2008 Kevin Karplus For try5, I'll start optimizing from the 4 chimeras, with high crossover, in the hope of picking up something good from each. Tue May 13 15:23:29 PDT 2008 Kevin Karplus try5-opt3 scores best with the try5 costfcn, and on align_sheets. try4-opt3 does a bit better on the Fe2S2 site. MQAC1 and MQAU1 do better on align_constraints, and have shifted less from the alignments. rosetta likes try5-opt3.gromacs0.repack-nonPC best, but that loses the Rieske Fe2S2 site. Perhaps I should do another optimization with breaks and soft_clashes turned up starting with just the gromacs-tweaked models. Tue May 13 17:15:11 PDT 2008 Kevin Karplus try6-opt3 scores best with the try6 cost function, and rosetta likes try6-opt3.gromacs0.repack-nonPC best of the ones it has repacked. gromacs removed 4 of the H-bonds, and rosetta removes 2 more, but the C80.O-K84.CA clash is reduced by gromacs. Since C80 is part of the rieske site, the constraints on it may be leading to try6-opt3 accepting the clash. I think I've reached the point of diminishing returns on this protein. I'll probably submit try6-opt3 MQAC1-opt3 MQAU1-opt3 try4-opt3 align1 (to 1uliA) Thu May 15 10:27:19 PDT 2008 SAM-T08-MQAO hand QA T0391 Submitted Thu May 15 10:27:19 PDT 2008 SAM-T08-MQAU hand QA T0391 Submitted Thu May 15 10:27:19 PDT 2008 SAM-T08-MQAC hand QA T0391 Submitted Sat May 24 18:39:46 PDT 2008 Kevin Karplus I'm going to do a try7 run with stricter clash definition. try7 < try6 < try5 < chimera-MQAC1-try4 Sat May 24 19:47:45 PDT 2008 Kevin Karplus try7-opt3 has slightly worse breaks than try6-opt3, but fewer clashes and better hbonds. Rosetta likes the gromacs-manipulated backbone better than the raw undertaker one. The residues it doesn't like in try7-opt3.repackPC are T16, K28, Y45, C57, C80, I99, F138, C140 It may not like the CYS, because it was not allowed to repack them. Y45.CD1 and K28.CA touch but undertaker thinks they're ok. C57 and C80 come close because of the Fe2S2 binding, but C80.O bumps K84.CA undertaker recognizes a self-bump in I99. C140.CB bumps S17.CB slightly. I don't think there is much point to further tweaking. Sat May 24 20:18:41 PDT 2008 Kevin Karplus I submitted ReadConformPDB T0391.try7-opt3.pdb ReadConformPDB T0391.MQAC1-opt3.pdb ReadConformPDB T0391.MQAU1-opt3.pdb ReadConformPDB T0391.try4-opt3.pdb InFilePrefix ReadConformPDB T0391.undertaker-align.pdb model 1 = T0391-1uliA-t06-local-str2+near-backbone-11-0.8+0.6+0.8-adpstyle5.a2m