Thu Jun 27 11:35:13 PDT 2002 Kevin Karplus Another clear homology-modeling target---this time to 2cpl (SCOP domain b.62.1.1). The highest scoring target with the w0.5 model is 2cpl, which happens to be already in the template library. Here are the top hits with the different models id w0.5 STR ALPHA DSSP SRIDE template RANK RANK RANK RANK RANK RANK 2cpl 1 2 3 2 1 2 1cwoA 2 1 1 1 2 6 1awqA 3 3 2 3 3 1 1awsA 4 4 4 4 4 10 1dywA 5 5 5 5 5 9 1ihgA 8 6 6 6 6 4 2rmcA 10 8 8 8 9 3 1a58 12 9 9 9 7 5 1qoiA 11 7 7 7 8 8 1lopA 15 10 10 10 10 7 The only disagreement seems to be whether to use 2cpl or 1cwoA or 1awqA as the template. According to FSSP, 2cpl and 1cwoA are within 0.2 Angstroms of each other over the full 164 residues and 2cpl and 1awqA are within 0.5 Angstroms), so it probably doesn't matter much which we chose as the template. Any differences in the template models are probably due more to when the models were built than to differences between the templates. We can tighten up the thresholds for choosing the best templates, to get only the SCOP domain b.62.1.1, since all the top hits are from that domain and there is a huge gap before any other domain. 28 June 2002 Kevin Karplus All the alignments included in T0144.undertaker-align.pdb.gz are almost identical---they superimpose almost perfectly. T00144.try1-opt looks great. It might be good to run it through SCWRL and do some rotamer optimization. Hmm, it also looks like there may be couple of disulphides: C40-C168 C69-C122 The C69-C122 pair, though close, are on opposite sides of a sheet and the 2cpl structure appears to have CYS on opposite sides of the sheet in a corresponding location---therefore most likely NOT a disulphide. Th C40-C168 pair DO look like a good candidate for a disulphide bond though, so let's try again with that added to the score function. 29 June 2002 Kevin Karplus T0144.try2-opt.pdb resulted in a bad break from InsertSSBond. I implemented a new optimization operation in undertaker (ImproveSSBond) and tried again with the same scoring function. The new best scorer, T0144.try3-opt.pdb, does not have quite as good an ssbond score as try2-opt, but has no breaks. I'll have to look at it to see how good it looks. 30 June 2002 Kevin Karplus RATS---the first strand (4-12) of the barrel has been pulled off and turned into a helix in try3-opt. Maybe I need to add constraints for the H-bonds: V6 M22 F8 I20 D9 A166 A11 T164 Doing so makes the try2-opt run score best of the current decoys (so it does correctly reject the helix formed in try3-opt). I'll try doing another optimization run with the new scoring function. I'll include the try1-opt, try2-opt, and try3-opt conformations as starting points, so that crossover operations have a chance of getting to a good conformation quickly. Hmmm, try4 seems to get trapped into variants of try2, and doesn't seem to be reducing the break much. I also goofed and forgot to rename the input conformations so that I could output intermediates to the decoys directory. Retrying after editing undertaker.script (still calling it try4). I just noticed that only one of the V6-M22 Hbonds was in try2-opt, and there was a bit of a bluge at 166-167, so that the pairing was A166:N,D9:O, D167:O,D9:N. This means that try2-opt will not score as well as it might with the new scoring function. It could also mean that I'm trying to flatten out a beta strand that is SUPPOSED to have a bulge. Despite the constraints on V6, it still manages to pull away from the barrel on try4. Perhaps I should loosen the constraints on D9 and tighten them on V6? Maybe remove the D9:N constraint entirely? I'll also turn off radius_fit, which seems to be misleading, and increase the penalty for breaks, so that try2-opt, with its bad break, does not score so well. try5-opt is looking quite good. There is an anomaly in the scoring, with score-decoys.rdb getting a different score from that in decoys/undertaker-try5.rdb, but there is a fairly good reason. The difference in scoring comes primarily from a difference in the scoring of the break before residue 66. Because the duplicated atoms are not output to the PDB file, undertaker has to guess how to complete the peptide planes at any break it finds. The way it guesses is not likely to reproduce the pre-existing break, and will often change the cost. The "break" here is quite small---the C of P65 is a little too close to the N of N66. I should probably make an attempt to improve the reconstruction of the peptide planes at breaks, so that scoring can be more consistent. 1 July 2002 Kevin Karplus try6-opt looks quite good. The surface is still a bit "foamy", with the residues not packed as tightly as one would like. Perhaps we should increase the weight for dry6.5 and maybe for gen6.5. At least try6-opt has finally gotten rid of the break. It might be good to try running SCWRL on try6-opt also. We'll have to install it for the LINUX boxes. 12 aug 2002 Kevin Karplus Reran make with the new template library. No change to top hits. I made sw-best to get the top single-sequence scores (1dywA is best), and made try-alignments. I modernized the score function and undertaker.script, and reran undertaker starting from just alignments, not existing conformations. Fri Aug 16 14:06:54 PDT 2002 Kevin Karplus New best score is T0144.try8-al8.3.60.pdb, which looks pretty good except for the bad breaks around C168-S172. I wonder how much of the problem comes from the SSBond S40-S168? Perhaps we should rescore (and reoptimize) without the bond. Date: Fri, 16 Aug 2002 14:33:00 -0700 From: Kevin Karplus To: karplus@soe.ucsc.edu, rachelk@soe.ucsc.edu, weber@soe.ucsc.edu, learithe@cats.ucsc.edu, yael@biology.ucsc.edu, baertsch@soe.ucsc.edu, rph@soe.ucsc.edu, afyfe@soe.ucsc.edu, jcasper@soe.ucsc.edu, oscarhur@soe.ucsc.edu Subject: SS bond? I have one protein that I'm not certain whether to include a disulphide bond in or not. The protein is T0144 and is supposedly a cyclophilin. We have a pretty good structure in T0144.try8-opt-scwrl (including the disulphide), but the disulphide is in the part of the structure that is worst. I'm trying a new optimization without the disulphide, but I could use some guidance on whether it makes sense biochemically to have it or not. Perhaps someone with more biology could tell me whether the molecule is usually extracellular or otherwise in an oxidizing environment, which would favor the inclusion of the disulphide. Date: Fri, 16 Aug 2002 17:27:45 -0700 (PDT) From: Jenny Draper To: Oscar Hur cc: Kevin Karplus , rachelk@soe.ucsc.edu, weber@soe.ucsc.edu, yael@biology.ucsc.edu, baertsch@soe.ucsc.edu, rph@soe.ucsc.edu, afyfe@soe.ucsc.edu, jcasper@soe.ucsc.edu Subject: Re: SS bond? >From what I can see by a bit of web browsing, cyclophilins tend to be intracellular. However, some are secreted (like cyclophlin B), and some are found in places like the mitochondrial matrix... -Jenny On Fri, 16 Aug 2002, Oscar Hur wrote: > > Well, I will give it a try. > > > S-S bond are not typically found in intracellular protein because they are > usually in reducing environment. S-S bond is usually in extracellular > protein that are secreted from cells. An example is insulin. Date: Fri, 16 Aug 2002 17:57:07 -0700 (PDT) From: Jenny Draper To: Oscar Hur cc: Kevin Karplus , rachelk@soe.ucsc.edu, yael@biology.ucsc.edu, jcasper@soe.ucsc.edu Subject: Re: SS bond? I found this, in the third link down in a google search for cyclosporin "disulphide bond" : X-ray Structures of Divergent Cyclophilins, J. Dornan, G. Kontopidis, P. Taylor, M.D. Walkinshaw "As in the cyclophilin from C. elegans, the two highly conservedcysteine sulphurs are well defined in the electron densityand are separated by 5.4 angstroms. A small adjustment in the sidechain torsion angles of both residues would enable adisulphide bond to form." Not sure if that helps -- it seems to be saying that in some cyclosporin structures, the cystines are very close, but still may not actually be forming the disulphide bond. Overall, it appears that disulphide bonds are not common in cyclophilins (probably explained by most cyclophilins being intracellular). 17 Aug 2002 Kevin Karplus try9-opt is new best, optimizing and scoring without the SSBond, but it forms the bond anyway, so let's put it back in the score function. try8-opt does slightly better with SSbond scoring on, but the difference is small. Let's do one optimization run starting from try8-opt and try9-opt with the SSbond on, then call it quits. 17 Aug 2002 Kevin Karplus try10-opt is new top score. It improved quite a bit on try8 and try9, so I think that another polishing run may be called for. This one will have stricter break penalties, and will turn up the wet6.5 weight to try unburying E83 and D73, which are covered by a loop. 18 Aug 2002 Kevin Karplus try11.16.80 is new best. It looks good, but D73 and E83 are still buried, and the structure is just a bit foamy. Let's turn wet6.5 down again, and see what scores best. The same structure is still judged best. Let's do moe more run from a couple of the best structures. 18 Aug 2002 Kevin Karplus try12.8.100 now best. Not much change from the try11 run, so let's just submit it.