Fri Jul 5 03:28:09 PDT 2002 5 July 2002 Kevin Karplus Most of the sequences found by t2k are DNA polymerases or "histidinol phosphatase and related hydrolases of the PHP family" with a scattering of other RNA and DNA-related enzymes. The main hits in PDB are oxidases and reductases from SCOP fold c.1 (TIM barrel). There are many TIM barrel superfamilies with phosphate-related functions, and this is not a surprising fold for such an enzyme. Since there are so many TIM-barrel hits, let's tighten up the BEST_EVALUE threshold to get down to 30 or fewer templates. The CAFASP servers generally agree that this is a TIM-barrel. 10 July 2002 Kevin Karplus PROBLEM: the alignments generally do not have all the strands of the barrel. I looked at the 1frb and 1gox alignments, and was NOT impressed. The results seem to be far too scattered, without contiguous beta sheets. Perhaps we need to look more at the template-model alignments. try1-opt.pdb doesn't have ANY of the beta strands clustered, though there is a hint that S161-L163 pairs with W186-A188, since they come out moderately close together. 11 July 2002 Kevin Karplus Model 41 T0147-1mdl-2track-protein-STR-local-adpstyle5 looks like half a TIM barrel. Maybe we should make up a double-length target and try to model it. 18 July 2002 Kevin Karplus I finally got around to looking at t0147_twice/decoys/T0147_twice.try1-opt.pdb It looks terrible---there are no beta sheets left at all. Even the inserted alignments don't look good---this approach seems not to be working. 11 Aug 2002 Kevin Karplus I reran make with the new template library---no real changes. I added the try-alignments target to the Makefile, but have not yet updated define-score.script or undertaker.script. 12 Aug 2002 Kevin Karplus Modernized define-score.script and undertaker.script and reran undertaker. Fri Aug 16 17:43:25 PDT 2002 Kevin Karplus try2-opt-scwrl scores a bit better, but still looks terrible. Maybe someone needs to look at the models in T0147.t2k.undertaker-align.pdb.gz and find appropriate H-bond constraints to add to the score function. In the meantime, for try3, I'll try starting from the robetta models. 17 Aug 2002 Kevin Karplus try3-opt-scwrl is new best. It seems to be based on robetta1. The first strand V4-M8 has gotten detached, but the rest looks ok. There seems to be some misalignment of I96-H101 and 71-85. Robetta puts hydrogen bonds for V4.N-A36.O, but I think that it should slide back 2 to L34. I'll have to work on these constraints, and think about getting a better alignment to start from. 18 Aug 2002 Kevin Karplus Let's try a run from both try2-opt-scwrl and try3-opt-scwrl, with CrossAndInsert set high, to see if we can salvage pieces from each. (I think we may need a more powerful crossover operator---one that takes a segment or subtree from one and applies it to the other.) 19 Aug 2002 Yael Working on the constraints to try get the full barrel, based on TIM beta/alpha barrel fold. Our ss predictions do not predict all the 8 strands of the barrel (should have a strand between each helix) so I am trying to look for the best alignment to help the strands form. 20 Aug 2002 Yael Adding constraints to get the 4 conserved histidins H9, H40, H131, H194 closer together... most probably involved in coordinating the metal binding. Expected to be in the c terminal end of the barrel. 20 Aug 2002 Yael Running try5 with constraints for forming a TIM barrel and proposed metal binding catalitic site described in try5.constraints. 21 Aug 2002 Yael Try5-opt scores the best but looks very poor very far from a TIM barrel. Dispite the constraints the histidine (9,40,131,194) (which I expect to be part of the active site) did not come close together and the strands do not form a barrel. Maybe we should put stronger weights on the constraints. 21 Aug 2002 Yael Added hbonds constraint for strand 2-3 5-6 6-7 based on alignments 1ezw (1-2) 1a4m (5-6, 6-7) changed the cost function for constraints to 0.05. Reading only 1frb, 1a4m and 1ezw alignments in InFilePrefix in which some of the strands are correctly aligned. Maybe next run we should read only the fragments of the alignments which corresponds to correctly aligned strands. Running try6... 22 Agu 2002 Yael try6-opt scores the best but still dosen't look too good and the barrel is not formed (though some strand are starting to form a sheet). Based on our discussion at the meeting today I will create fragment alignments for the parts of the sequence which aligns well (from different parents) and read them to undertaker (before TryAllAlign). I will also add constraints for the last sheet which to not predict in our ss predictions for to close the barrel. I will also add stronger constraints for getting the histidines closer together. From yael@darwin.UCSC.EDU Mon Aug 26 10:38:04 2002 MIME-Version: 1.0 X-Authentication-Warning: darwin.UCSC.EDU: yael owned process doing -bs Date: Mon, 26 Aug 2002 10:37:12 -0700 (PDT) From: yael@biology.ucsc.edu To: Kevin Karplus Subject: Re: editing alignments In-Reply-To: <200208240135.g7O1ZXp30835@bray.cse.ucsc.edu> Content-Type: TEXT/PLAIN; charset=US-ASCII X-SpamBouncer: 1.4 (10/07/01) X-SBPass: NoBounce X-SBClass: OK Try7 doesn't score good (try6 is still the best) and doesn't look like a barrel at all and the histidines don't look in place either. It does seem that the constraints did get some of strand pairs together. I looked at the log file and one of the hand alignment couldn't be read because one aa that was misaligned. I will fix it now. I wonder if I should try and run try8 without constraints now or without TryAllAlign. Before I do it I will put all my remarks in the README file Yael Yael Mandel-Gutfreund Department of Chemistry and Biochemistry University of California, Santa Cruz Phone : 1-831-459-5128 e-mail : yael@darwin.ucsc.edu On Fri, 23 Aug 2002, Kevin Karplus wrote: > You are correct, to use the new alignments you probably want to use > TryAllalign. > > I edited the undertaker.script for you and started the try7 run. > It should finish sometime tonight, for you to look at in the morning. > > Kevin > > Aug 26 2002 Yael Fixed the hand alignment. Should we run try8 again without TryAllalign? Mon Aug 26 15:39:55 PDT 2002 Kevin Karplus Yael says she's fixed the hand alignment---I assume she means T0147-1ezwA-hand.a2m.gz I'll edit undertaker.script to try using it, and run a short try8, to see if anything comes up. I've modified the score function to pay a bit more attention to the constraints and to burial. The best-scoring one is currently try6-opt-scwrl, and it IS beginning to look like a barrel. There seems to be a hairpin coming out of the barrel, with E82-C85 paired with (probably) A74-K77. These are not in try8.constraints---let's add them to try9.constraints. It looks to me like there are enough histidines to have TWO metal-binding sites. High conservation for H7 H9 H40 H131 H194, low conservation for H101. All six of these come close together in try6-opt-scwrl. Let's group them in two sets of 3: H7, H9, H40 H101, H133, H194 Note: an alternative triple would be H7 H40 H101 or H7 H131 H194. Hmm, there are many possibilities. Maybe I should loosen these constraints quite a bit, so that they don't pull the barrel out of shape. I wonder if the conserved E73 and D192 are also involved in the active site? They seem to be close to the histidines also. I'll try running a short script with the try9.constraints, polishing from best-scoring models from tries 3 through 7. 26 August 21:28 The best score is now for try9-opt, but try8-opt-scwrl is not too bad, being beaten by try6-opt-scwrl and several of the try9 iterations based on try6-opt-scwrl. Both try9-opt and try8-opt-scwrl have messed up the barrel, so I suspect some of the constraints are wrong. One of the reason we are having so much trouble is that TIM-barrels have 8 strands, but we've only predicted 7. Finding the other strand to close the barrel is tough. (Cellulases, scop domain c.6, have only 7 strands, but none of our hits are to a cellulase domain) There must be some slippage from strand to strand around the barrel, to get closure. Normally a TIM barrel slips by 1 on each strand +++++ +++++ +++++ +++++ +++++ +++++ +++++ +++++ Let's try to get closure with only 7 strands. 4 5 6 7 8 34 35 36 37 38 66 67 68 69 70 95 96 97 98 99 126 127 128 129 130 160 161 162 163 164 186 187 188 189 190 4 5 6 7 8 mypVDLHMht ikLFAITdh gVGILRgie dLIIAGfh nVHIIShp nSSFLHsr gWVALGsdsh mypVDLHMhtv ikLFAITdh To keep all the conserved H on the same face of the sheet, we need to make strand6-7 or strand7-1 slip by 0 or 2 instead of 1. Since we have high conservation of D192, let's make the slip in 6-7. Note that the highly conserved E73 also lined up with the highly conserved H on the inside of the barrel. Let's conjecture E73, H101, H130, H163, D192, H194, H7, H9, and H40 all come close on the inside of the barrel. Since these constraints (try10.constraints) are so different from Yael's I'll probalby have to run undertaker from scratch. The best of the current models is try9-opt-scwrl, but it hasn't done a good job with the barrel. I'll run try10 overnight and see what it comes up with. Date: Mon, 26 Aug 2002 23:31:47 -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: T0147 I've taken t0147 over from Yael, as she has started having contractions---the birth definitely takes precedence over CASP! I hope the labor goes smoothly for her. In any case, I just started looking at the file today, but I think I've come up with a nice set of constraints that match all the information that Yael provided. All the conserved residues seem like they should cluster together in the center of the barrel near one end. The constraints Yael was using were trying to achieve this effect, but I think she forgot that a barrel has to slip a residue with each strand in order to get the right slant to the sheet, so she was having a hard time getting a consistent barrel. Also, she was trying to do an 8-strand TIM barrel, but we only have 7 predicted strands. I considered trying to find another strand in there somewhere, but eventually gave up. Instead I'm trying to make a 7-strand barrel. (There is some precedent--the cellulases in fold c.6---but I wouold have been happier with a TIM barrel.) Take a look at the README file to see the barrel I'm trying to get, and send me any comments on ways to improve it. This is due Wed 28 Aug, so I don't have much time to work on it. Date: Tue, 27 Aug 2002 00:13:11 -0700 (PDT) From: yael@biology.ucsc.edu To: Kevin Karplus Subject: Re: T0147 Thanks Kevin, As you can understand I am still around though I don't think it will be much longer. In any case I wanted to refer (who ever is going to work on t0147) to a paper I found today on the PHP motif which is expected to be the catalytic site of t0147. Phosphoesterase domain associated with DNA polymerase of diverse origins Aravind and Koonin NAR 1998 26(16) 3746-3752 (it is available on line) The paper gives some clues about the involvement of the histidines in the catalytic (metal-dependent) reaction. Though it is not clear exactly how, they do predict that two metals are probably involved and coordinated by the conserved histidines. The conserved D is probably also involved in the catalytic reaction (it could be different in the different superfamily). Another important information from this paper is that they find a weak hit for the PHP superfamily with 2kau which is a disordered TIM barrel fold. I haven't managed to look at 2kau but I think it is worth while to do. If I am still around tomorrow I will look at it. Good luck Yael Yael Mandel-Gutfreund Department of Chemistry and Biochemistry University of California, Santa Cruz Phone : 1-831-459-5128 e-mail : yael@darwin.ucsc.edu ------------------------------------------------------------ 27 Aug 2002 01:21 Kevin Karplus The barrel is in 2kauC (not 2kauB or 2kauA). We have 2kauC in our library, also the 64% identical sequence 1ubpC, which agrees to within 0.9 Ang over 566 residues. This is SCOP domain c.1.9.2, and several of our top hits are to this domain---including 1ubpC. So we have already been checking out these possibilities. I'll be more explicit about them in try11, trying only the c.1.9.2 domains in the TryAllAlign, and omitting the barrel-closing constraints in try11.constraints. 27 Aug 2002 08:29 Kevin Karplus try10-opt has gotten close, but not close enough. It is trying to knot the protein around D171---I'll have to move V176 to the end out of the way and repack. I might also want to check the register and redefine some of the constraints? try10-opt has L6.O A36.N OK f35.O G71.N different register strand3 +3 I37.N G71.O I37.O G73.N I72.N G99.O OK register, but past end of predicted strand I72.O H101.N F100.N H127.O different register strand5 -2 F100.O I129.N I128.O E156.N different register strand6 -6 S130.N E156.O What did I want as the register? oioioioioioioioioio mypVDLHMht ikLFAITdh gVGILRgie dLIIAGfh nVHIIShp einnSSFLHsrk gWVALGsdsh mypVDLHMhtv ikLFAITdh Hmm---strand 2-3 constraints are WRONG in try10 and try11.constraints. I'll start try12 trying from scratch from the alignments, and try13 starting from the best existing models (plus from alignments). (try13 started when try11 finished, around 10:14---for try13 I loosened the constraints on the active site residues, as they seemed to be causing problems for the sheet formation.) I also upped the break and clash penalties, to try to get more coherent pieces. 27 Aug 2002 Kevin Karplus 15:16 New best is try13-opt. Best from the try12 run is try12-opt-scwrl. In try13-opt,strands I96-H101 and G71-K77 are out of place. Are the constraints wrong? 17:29 I think that the constraints for clustering the active site residues may have been too strong, so I'll try reoptimizing without them (try14). There also may be some register problem that puts some of the active-site residues on the wrong side of a strand---I'd better double-check. 18:40 try14-opt is looking better, but the N terminal strand and the one with the hairpin are trying to form a knot with the C-terminal end of the chain. Let's add a few constraints to help out---maye H164-H7, H7-H194, H194-H9. Also, we can move V66-D94 way out (say in the direction of S123) and let it move back with OptSubtree. 27 Aug 2002 21:12 Best score is now try15-opt-scwrl Let's see what constraints are needed: H164-H131-H101-H40-H9-H194-H7-H164 D192 should be centered on top, E73 on the bottom (maybe 9 Ang apart?) 28 Aug 2002 060:03 Kevin Karplus Best score now try16-opt. The strands involving I68 and G99 are twisted around each other. No part of V66-R70 seems to be part of the active site, so is not heavily constrained---we may need to add more constraints to hold it in place. The break at L34-F35 looks like it will require moving S18-L34 so that Q26 is on the other side of F197. It looks like the hairpin should fold over the C-terminal end of the barrel, to position E73 over the cetner of the metal ion---on the same side as D192. Let's move R70-G87 up (towards D46) so that it can be repositioned. 08:51 Best now try17-opt Try changing HIS order to H9-H40-H101-H131-H164-H194-H7-H9 Date: Wed, 28 Aug 2002 10:40:26 -0700 (PDT) From: yael@biology.ucsc.edu To: Kevin Karplus Subject: Re: team meeting Wed 11am Sorry, I will not be able to come to the meeting today (have an OB appointment and may have to stay in hospital after) I managed to look at T0147 and T0148 T0149 (mainly at the opt files) Here are my suggestions for models to submitt (ranked by order): T0147 1. try14-opt 2. try17-opt 3. try13-opt ---------------------------------------- Wed Aug 28 15:02:05 PDT 2002 Kevin Karplus Maybe I need to modify the HIS constraints---make the adjacent H residues be closer and loosen the constraints on D and E. Wed Aug 28 15:55:04 PDT 2002 I just noticed that the entire try18 run had both the try17 and the try18 constraints! Here is current mapping of where conserved residues are on intended ot be on barrel: H9-H40-H101-H131-H164-H194-H7-H9 E73 D192 m y p\v d l H m\H t k g i k\l f a i t\d H g\v g i l r\g i E d\l i i a g\f H e p n\v h i i s\H p n\s s f l H\s r g w\v a l g s\D s H m y p\v d l H m\H t v i k\l f a i t\d Perhaps the 4th strand needs to be back off 2: m y p\v d l H m\H t k g i k\l f a i t\d H g\v g i l r\g i E d l i\i a g f H\e p n\v h i i s\H p n\s s f l H\s r g w\v a l g s\D s H m y p\v d l H m\H t v i k\l f a i t\d I've put these constraints into try19.constraints. Wed Aug 28 17:26:38 PDT 2002 The first iteration, T0147.try19-try18.0.80.pdb, is looking better. I hope this manages to get the strands into place. It seems to be relying mainly on Insertfragment to move things around. The E73 is NOT flopping over onto the active site, and so is twisting the preceding strand out of place. Perhaps I should try adding a "bending" constraint aroung G71 and tighten the constraints for the perceding strand. The D192 and H194 are also having a hard time getting to the inside of the barrel---perhaps I should back them up a bit, though this violates the prediction of the position for beta strand 186:GWVALG m y p\v d l H m\H t k g i k\l f a i t\d H g\v g i l r\g i E d l i\i a g f H\e p n\v h i i s\H p n\s s f l H\s r g w v a\l g s D s\H m y p\v d l H m\H t v i k\l f a i t\d This is try20.constraints. Wed Aug 28 19:55:39 PDT 2002 Even with the active site constraints removed (try20-noactive.constraints), try20-opt scores best. try20-opt has sort-of formed the barrel (almost no Hbonds, but the strands are roughly in the right places). With the try19 constraints, try19-opt-scwrl scores best. The active sites of try20-opt and try19-opt-scwrl look about equally bad. With the try18 constraints, try18-opt-scwrl scores best. Again the barrel is sort of there, and the active site looks about the same. With just the active-site constraints, try9-try6.4.80 scores best. This has a pretty good barrel, and the best active site I've seen. Maybe I should do a quick polishing run with constraints turned down and just the active site constraints? With no constraints, try9-try6.2.80 scores best. The active site looks ok (not much worse than the others), and this actually has a pretty good barrel also. Wed Aug 28 21:24:36 PDT 2002 1: try21-opt 2: try20-opt 3: try19-opt-scwrl 4: try9-try6.2.80