Sun Aug 27 14:22:38 PDT 2000 Kevin Karplus >T0123 Beta-lactoglobulin, pig Excellent blast hits to 1bsoA, 2blg, 1bsy, 3blg, 1dv9A, 1cj5A, ... Also excellent double-blast hits. TARGET HMM SCORE INTERE FINALE intermediate 30S 30S 10N 10N 10N 26S T0123 1cj5A -120.049 5.4e-80 7.3e-53 gi|346460|pir||A45542_19:178 T0123 2blg -120.049 5.4e-80 7.3e-53 gi|346460|pir||A45542_19:178 T0123 3blg -120.049 5.4e-80 7.3e-53 gi|346460|pir||A45542_19:178 T0123 1bsoA -120.049 5.4e-80 7.3e-53 gi|346460|pir||A45542_19:178 T0123 1dv9A -120.049 5.4e-80 7.3e-53 gi|346460|pir||A45542_19:178 T0123 1bsy -120.049 5.4e-80 7.3e-53 gi|346460|pir||A45542_19:178 T0123 1bebA -118.348 5.4e-80 4e-52 gi|346460|pir||A45542_19:178 T0123 1bebB -118.348 5.4e-80 4e-52 gi|346460|pir||A45542_19:178 T0123 1b0o -117.337 5.4e-80 1.1e-51 gi|346460|pir||A45542_19:178 T0123 1bsqA -117.337 5.4e-80 1.1e-51 gi|346460|pir||A45542_19:178 183 sequences in t2k alignment, 46 of which are in PDB. The family that contains T0123 and some PDB files is T0123-family.a2m.gz The PDB files have 4 distinct sequences: 1 1bsoA, 1bsy, 2blg, 3blg (note: structure depends on pH) 2 1cj5A, 1dv9A 3 1bebA, 1bebB 4 1bsqA, 1b0O All are Bovine Beta-lactoglobulin. pH is given as 3.2, so we should probably look for the most acid of the structures: pH 1bsoA 7.3 1bsy 7.1 2blg 8.2 3blg 6.2 1cj5A 2.0 1dv9A 2.7 1beb[AB] 6.5 1bsqA 7.1 1b0o 7.5 It looks like 1cj5A and 1dv9A are the appropriate ones to use as a model. Both are represented in FSSP by 1bebA, but with poor fits (2.5 and 1.4 Angstroms). I'll build basic setups for both 1cj5A and 1dv9A. I don't see much difference between them by eye. Both are NMR models, and the 1dv9A files says that the best representative confomer is number 14---maybe we should build a model using that confomer and running scwrl. I have created such a scwrl file as 1dv9A/T0123-family-1dv9A-14-scwrl.pdb Unless I get a chance to use undertaker to close the small gap near the C terminus, we should probably go with that. For secondary structure prediction, the T0123.t2k.2d prediction is MUCH better than the T0123-family.2d prediction. Tue Sep 5 10:35:25 PDT 2000 remaking 2track 8 Sept 2000 Kevin Karplus First undertaker run screwed up, because initial conformation not read from right file. Note: t123 does use standard residue numbering. Run try2 is ok, but didn't close gap. Undertaker run try3 is ok, but is not closing the gap. Undertaker run try4 is ok, but is not closing the gap. Let's try starting just by applying the two fragments separately, and not using the 1dv9A/T0123-family-1dv9A-14-scwrl.pdb starting conformation. Run try5 is finally getting a good result---the second disulphide bridge is formed in t123-opt.5.10.pdb, which looks very good. Maybe I should try again with the SSBOND commands to get better scoring? Perhaps starting from the best structure so far found? Sun Sep 10 10:41:32 PDT 2000 Re-SCWRLing from try5/t123-opt.pdb with known SSBonds in scoring function produces slightly better results: t123-init at pool[3] 22.7996 bits/residue, 12 clashes t123-init-scwrl at pool[4] 20.8338 bits/residue, 14 clashes and still has two SSbonds. Reoptimizing actually loses one of the ssbonds---perhaps because the known SSbond weight is too low, perhaps because the softening of the ssbond scoring peak was too extreme. Trying again with larger known_ss_bond_bonus (20 instead of 1). try7 keeps the ssbonds, and manages to reduce the clashes from 14 to 7, though many of the superiterations stayed at the starting conformation. Optimizing the rotamers increased the number of clashes up to 17, and SCWRLing left it at 12. Let's submit try7/t123-opt.pdb