Mon Jun 21 07:21:53 PDT 2004 T0208 DUE 13 July Mon Jun 21 09:15:48 PDT 2004 Kevin Karplus The t04 alignment finds a very high similarity to 1i60A (SCOP c.1.15.4). The t2k alignment also finds this similarity, but doesn't regard it as quite so close---still close enough to be considered very good hit. It can't be full-length though, so we may have to split into domains. The 2-track searches are also finding 1i60A as an obvious top hit. According to PDB, 1i60 is a protein of unknown function, but the crystal data seems to be good (1.6 Angstrom). There are several converved residues, giving a strong signal. In the t2k alignment, I notice 3 conserved HIS and 1 conserved CYS, suggesting a metal-binding site for H207, C245, H272, H318. The t2k and t04 multiple alignments are not in agreement about the conserved residues, so the intersection of their conserved residues is probably the strongest signal for conservation. Mon Jun 21 11:05:05 PDT 2004 Kevin Karplus I looked at the alignments before any optimization. It looks like H207, H272, and C245 are part of a binding site. It isn't entirely clear whether R274 is part of same site. H318 is nearby, but currently on the other side of the beta sheet. The problem is that both D317 and H318 are conserved polars, but only one side of the strand is wet. The template barrel 1i60A looks a fairly standard TIM barrel---there are a lot of HIS residues, but they don't cluster in a point, instead several form a line leading down into the barrel. The alignments are also not getting the hbonds of the barrel. Why not? This is supposed to be a direct copy of the backbones from the template! Hmm, rasmol doesn't see any hbonds in 1i60A--why not? The structure claims to be 1.60 Angstrom resolution X-ray file. The N-O spacings seem to be too large sometimes, but others seem fine---why can't rasmol find them? I looked at the what-check report on http://www.cmbi.kun.nl/gv/pdbreport/ for 1i60, and there don't seem to be major red flags. Mon Jun 21 15:32:53 PDT 2004 Kevin Karplus try1 finished, and it almost has a TIM barrel. There are only 7 of the needed 8 strands, and the first strand is being formed by a region predicted to be helix. The 8 major strands are probably 1: W3-Y8 2: A60-R65 3: S100-S105 4: G153-G158 (a bit dubious) 5: K203-H207 6: G241-C245 7: N269-F274 8: G319-D324 I think that Y104 (a conserved polar) is on the wrong side of the barrel in try1---it should be inside. Y310 may have the same problem. We can try pasting in the first strand 317: dhgrAiwdek 1: mkWGFRWYgaa 58: glALLGIESva 99: gisLVCYSfkp 150: qskgfrlpgw 200: eniKMGIhp 238: pangITFCtg 267: riNFVHFRNVKYlg 317: dhgRAiwdek I'm pretty sure about pairing the conserved H207 and C245 and probably the H272. From karplus@soe.ucsc.edu Tue Jun 29 10:57:32 2004 Date: Tue, 29 Jun 2004 10:57:31 -0700 From: Kevin Karplus To: oscarhur@soe.ucsc.edu CC: karplus@soe.ucsc.edu In-reply-to: (message from Oscar Hur on Tue, 29 Jun 2004 04:16:45 -0700 (PDT)) Subject: Re: TIM barrel We do need better alignments and better strand pairing for the probable TIM barrel T0208. If you are interested in working on it, read the casp6/T0208/README and casp6/README files for information on what we are doing, and come to the Thursday meeting. Kevin ---------------------------------------- Tue Jul 6 08:16:30 PDT 2004 Kevin Karplus In try3, the strands are gathered into a rough barrel, but none of the hbonds are being made---we probably have play with the phase of the strands some more. From learithe@soe.ucsc.edu Thu Jul 8 12:54:03 2004 Date: Thu, 8 Jul 2004 12:53:57 -0700 (PDT) From: Jenny Draper To: Kevin Karplus cc: sol@soe.ucsc.edu, ggshack@soe.ucsc.edu, learithe@soe.ucsc.edu, martina@soe.ucsc.edu, bbarnes@ucsc.edu, marcias@ucsc.edu, rph@soe.ucsc.edu Subject: Re: please update README files Hi Kevin, As you have noticed, I haven't been able to do anything productive on T208 yet. The last two days have been mired in a bureaucratic health-insurance nightmare. I've spent a little time playing with hand-editing of the T0211 structure (Tuesday afternoon) and time looking at/researching T208. I've got appointments all day today (this morning, and 3:30-5:30 at Stanford, which amounts to 1:30-7:00 in travel time). So I won't be able to get anything done until tonight, so if you could look into it today, that would be really great. Have you looked at the Robetta models? I kind of like some of what I'm seeing, although they look pretty bad under the burial script. They all want to turn the region around 125-130 into a strand (it's weakly helical in our str2 predictions, and mildly strand with stride). They also want to break up the protein into two domains. Notation on the CASP6 known functional information website lists this as a "Mannonate dehydratase" http://cassandra.bio.uniroma1.it/Casp6/Targetinfo.html#_t0208 which can bind metal, specifically iron: Dreyer JL. The role of iron in the activation of mannonic and altronic acid hydratases, two Fe-requiring hydro-lyases. Eur J Biochem. 1987 Aug 3;166(3):623-30. Robetta's T0208 result/summary page seems to like "1c7nA1" for domain1: Ginzu Domain Prediction 1 (click on domain number for details) Domain Span Source Parent Parent Span Confidence Annotations 1 1-87 3djury 1c7nA_ 140-223 1.340000 TRANSFERASE 2 88-357 pdbblast 1k77A_ 18-250 33.000000 STRUCTURAL GENOMICS, UNKNOWN FUNCTION Could you take a look at this sometime today, and let me know what you think? Should I pursue this two-domain idea? If so, could you set it up? It would certainly save me some time trying to figure it out. I'm sorry about not getting anything done; I should have warned you sooner. I will be working over the weekend. -Jenny ============================================================ Thu Jul 8 21:11:20 PDT 2004 Kevin Karplus I think I'd rather pursue this as a single TIM-barrel, rather than as a 2-domain protein. We may be wrong, but it would take us a lot longer to make 2 domains, and I'm not impressed with the robetta results---some of the strand pairs look good, but some of the strands of the barrel are made with residues predicted to be helices. It looks like the mutual-information constraints may be useful for this one, even though George's neural net is currently not reporting any constraints. I'll put try4 together making sure the sheet constraints are compatible with sheets that are forming in try3-opt2. I'll try tweaking the alignments a bit to see if I can get better closure of the barrel. I'll start with alignments again for this one, but we may have to switch over to starting with existing models and doing lots of cross over for try5. Fri Jul 9 08:29:43 PDT 2004 Kevin Karplus I don't like try4 very much. The strand with Y104 on it seems to be oriented the wrong way (I'd like Y104 in the center of the barrel). E161-Q166, which I predicted to be helix, is turned into a strand. Also G28-L34, which is predicted to be helix has become a straight piece, though not one of the barrel strands. The last 4 strands look fairly good though. The C-terminal helix is pointing out into space, rather than lying down against the barrel, but A337 and L340 are predicted to be buried. Maybe we should add a constraint for them with I314 and W3. While I'm still convinced this is a TIM barrel, the alignments look very difficult for the first half. For try5 we should probably remove some of the more speculative constraints, which might be pulling the model in a bad direction. Fri Jul 9 5:30pm Jenny Draper Started try5 as a "why not, just for fun" run. It's creating a new conformation from alignments, using the try3 constfunction, but with sheet constraints added from try4-opt2.sheets. Also, I read in all of the alignments for the templates 1k77A and 1i60A. Fri Jul 9 8:15pm Jenny Draper Started try6, which is an alignment run using the full set from 1k77A and 1i60A (like try5). This time I'm using a cleaned-up version of try4.costfcn. I removed all the non-strand related predicted constraints/contacts (except for the helix definitions, and a 245-319-ish pair that keeps showing up). I figure this should give undertaker some more "room" to form the barrel. Other changes include: * a bunch of atom constraints from try4-opt2.constraints from the nice looking bit of sheet in try4-opt2 * attempt to replace the strand 153-158 with a strand from 126-129. I'm trying this because: - position of the strand in the region 110-160 seems pretty much up in the air - this region is helical in STR2, strand in STRIDE, EHL2 - this strand (currently 153-158) seems to be a major problem area in the current models. none of our alignments match this region. - robetta's models chose to use this strand, and they get a decent barrel structure (though Kevin doesn't like the models overall) I have no idea if this is remotely the right thing to do, but I figured I'd give it a try I'll start try7 as an optimization, once I get the results of try5 & try6. So far, try5 looks like a blend of try3 & try4. Like try4, it's put helix 161-173 into the barrel. Ick. Fri Jul 9 9:15pm Jenny Draper Try5 is done. Try3 still scores better. Try5-opt2 has a decent sheet forming -- except that it's got helix 161-173 embedded right in the middle.... bummer Sat Jul 10 12:15pm Jenny Draper Try6 is done. It holds the sheet from try4 well, but otherwise is quite a mess. I'm going to start try7 from the try3 conformation, using a version of the try6 cost-function that makes the strand and helix constraints have a bit more weight. I'm also going to start try8 from all of the conformations (try3,4,5,6). Sat Jul 10 7:00pm Jenny Draper Something happened, and my runs got killed partway through. Try8-opt1 was done; I'll rename it to try8.5-opt1.... Re-starting the runs now... Sat Jul 10 10:00pm Jenny Draper It happened again. Apparantly the guys have been messing with the ethernet hookups, and keep killing my ssh connections -- and for some reason, this kills my undertaker runs, even though I have them in the background. So... it's like I've got nothing done today. Dammit. ... trying again... Sun Jul 11 2:00pm Jenny Draper Well, try8 finished, but try7 died after producing opt1. I have no idea why? The log file just stops, no error message. Try 8 and try7-opt1 look terrible. Sun Jul 11 11:00pm Jenny Draper I'm starting try9 from a hand-made barrel, which I put together in DeepView from try3-opt, using (very roughly) these strands: > >The 8 major strands are probably > 1: W3-Y8 > 2: A60-R65 > 3: S100-S105 > 4: L126-E130 (was G153-G158) (a bit dubious) > 5: K203-H207 > 6: G241-C245 > 7: N269-F274 > 8: G319-D324 This structure is called T0208-try3-opt2-hand4.pdb. Try9 is using the try7/try8 constraints. Mon Jul 12 9:30pm Jenny Draper The try9 run failed early, and I didn't catch it because I thought it would give me a "died" message if it failed; but apparantly when I run stuff in the background, it'll only show that message when I'm actively working in that terminal (like hitting return a couple of times). Dammit! Starting again... Mon Jul 12 19:19:10 PDT 2004 Kevin Karplus Trying to make sense of what is out there, I made a "strands" script: define s1 1-8 define s2 60-66 define s3 100-104 define s4 125-129 define s4a 153-156 define s5 203-207 define s6 242-245 define s7 270-274 define s7a 284-287 define s8 311-315 define s8a 319-324 Looking for pairings in existing models: s1-s2: Hbond G63 || W7 Hbond L62 || F5 s2-s3: Hbond S100 || L61 Hbond C103 || G63 Hbond L62 || L101 Hbond L101 || I64 s3-s4: Hbond S127 || L101 Hbond L128 || C103 s3-s4a Hbond Y104 || F154 s4a-s5 Hbond F154 || M204 s5-s6: Hbond I206 || I242 repeatedly Hbond I242 || H207 s6-s7: Hbond F244 || V271 repeatedly s7-s8: Hbond V313 || F270 repeatedly s8-s1 none Trying to build a TIM barrel using this and trying to keep conserved polars on the same face of the sheet: s1 mKWGFRWYg s2 ALLGIES s3 iSLVCYs s4a kgFRLp s5 iKMGIHp s6 gITFCtg s7 NFVHFRNVK s8 eGVIRpd s1 mKWGFRWYg SheetConstraint W3 Y8 A60 E65 hbond F5 SheetConstraint L62 S66 S100 Y104 hbond L62 SheetConstraint V102 Y104 F154 L156 hbond V102 SheetConstraint F154 L156 M204 I206 hbond F154 SheetConstraint I206 H207 I242 T243 hbond I206 SheetConstraint I242 C245 N269 H272 hbond I242 SheetConstraint N269 H272 G312 R315 hbond N269 SheetConstraint G312 R315 M1 G4 hbond G312 For try10, I'll put in these sheet constraints, and not worry about the helices or other problems. With these constraints, the try3-opt2 model looks best. I'll try reoptimizing using all existing models. From karplus@soe.ucsc.edu Mon Jul 12 20:50:16 2004 Date: Mon, 12 Jul 2004 20:50:14 -0700 From: Kevin Karplus To: learithe@soe.ucsc.edu CC: karplus@soe.ucsc.edu Subject: T0208 status I looked at T0208.try9-opt1, and it looks a bit better than try8, though try8-opt2 actually scores better with the try9 cost function---mainly because of lower clash and break costs, but also better burial. -------------------------------------------------- From learithe@soe.ucsc.edu Mon Jul 12 20:51:00 2004 MIME-Version: 1.0 Date: Mon, 12 Jul 2004 20:50:59 -0700 (PDT) From: Jenny Draper To: Kevin Karplus Subject: Re: status of T0208 In-Reply-To: <200407130038.i6D0c1wZ015988@cheep.cse.ucsc.edu> I think I like try3 > try2 > try4 > try5=6=7.5=8 -Jenny ================================================================= On Mon, 12 Jul 2004, Kevin Karplus wrote: > > Can you give me a quick status update on T0208? > > What are the best 3-5 models? > > I'd like to do a submission tonight. We can replace the models > tomorrow morning if something better comes up overnight, but I don't > want to risk having NO submission if there is a network failure tomorrow. > > Kevin > ------------------------------------------------------------ Mon Jul 12 21:12:15 PDT 2004 Kevin Karplus I've submitted try3-opt2 to vast: job id VS59743 password T0208try3 If we can get a decent alignment to some template, then using that alignment may give us the sheets! ------------------------------------------------------------ From learithe@soe.ucsc.edu Mon Jul 12 21:00:41 2004 MIME-Version: 1.0 Date: Mon, 12 Jul 2004 21:00:40 -0700 (PDT) From: Jenny Draper To: Kevin Karplus Subject: Re: T0208 status In-Reply-To: <200407130350.i6D3oE1J017747@cheep.cse.ucsc.edu> Right now, I'm working on a superpositioning from some of the best models. ------------------------------------------------------------ Mon Jul 12 22:28:57 PDT 2004 Kevin Karplus I've picked up two alignments from VAST: one for 1gw9A (actually for 1xif, which is the same sequence), the other for 1a0dA, which was not previously a template. I'll use both of these as starting points for try11, using the try10 costfcn. In doing the initial "TryAllAlign", undertaker prefers the 1a0dA alignment, probably because it is longer and so fewer residues are left in completely random conformation. I looked at the alignment in Cn3D, and it seems like the highly conserved residues do all seem to point to an Mn ion in the 1a0dA structure. This is very encouraging. ----------------------------------------------------------------- Mon Jul 12 11:00pm Jenny Draper I created a superposition of try2-opt2 1-200 try4-opt2 199-357 as decoys/T0208-try4-try2-super.pdb The first strand is pulled out of the sheet and facing out, and there are some breaks, but it looks better than any of the individual models so far. It may need un-DOS-ing... ---------------------------------------------------------------- Mon Jul 12 22:50:02 PDT 2004 Kevin Karplus I'm submitting try9-opt1 to vast also job VS59745 password T0208try9 When vast is done, I'll try doing an undertaker run with the best few alignments it has also, using the try9 cost function, which I have to assume Jenny has set up with a consistent set of sheet alignments. Hmm, maybe I should check that: SheetConstraint R315 D317 G4 R6 4 SheetConstraint F5 Y8 L62 E65 3 SheetConstraint L62 E65 V102 S105 5 SheetConstraint V102 S104 G126 L129 5 #higher weight SheetConstraint G126 L129 K203 H207 5 #higher weight SheetConstraint K203 H207 A239 T243 4 SheetConstraint I242 T246 N269 F273 4 SheetConstraint F270 F273 V313 P316 4 SheetConstraint F270 F273 V313 P316 hbond V271 SheetConstraint S100 V102 G124 L126 hbond V102 #ok SheetConstraint K203 H207 A239 T243 hbond M204 #ok SheetConstraint T243 C245 F270 H272 hbond F244 #good SheetConstraint V271 F273 I314 P316 hbond V271 #good s1 mkwgFRWYg s2 alLGIEs s3 iSLVCYS s4 GSLSLlf s5 iKMGIHp s6 ANGITFCTg s7 NFVHF s8 egVIRPD s1 mkwGFRWYg This seems to be off by one on the closure to get a normal slip: s1 mkWGFRWYg I think that the extra residue needs to be put in between S3 and S5, to keep the polars all on the same face: s1 mkwgFRWYg s2 alLGIEs s3 iSLVCYS s4 GSLSLLf s5 iKMGIHp s6 ANGITFCTg s7 NFVHF s8 egVIRPD s1 mkwgFRWYg There are also errors in the atom naming for try9: # HelixConstraint Error: residue specified as R151 doesn't match (T0208)S151 # HelixConstraint Error: residue specified as A298 doesn't match (T0208)M298 # StrandConstraint Error: residue specified as G126 doesn't match (T0208)L126 # SheetConstraint Error: residue specified as S104 doesn't match (T0208)Y104 # SheetConstraint Error: residue specified as G126 doesn't match (T0208)L126 I'll put together a new set of constraints that fix most of these problems, and use it for the run starting from the vast alignment to the try9 model. ---------------------------------------- Mon Jul 12 23:22:48 PDT 2004 Kevin Karplus The try11 run may be pretty useless, because the try3 alignment I started from did not have the strands where I wanted them---it had the helix in the strand position! No wonder the alignments with xylose isomerase were so good---this was essentially copied from alignments to other xylose isomerase crystals! Still, the try11-opt1 run scores better than any others (so far) on the try10/try11 costfcn, though try10 will probably beat it when it finishes, since it is going into the first super_pool with already better scores. ----------------------------------------------------------------- Mon Jul 12 12:00pm Jenny Draper I've messed with my superposition structure, to stick strand1 down into the helix... decoys/T0208-try4-try2-super-hand.pdb ------------------------------ Mon Jul 12 23:56:34 PDT 2004 Kevin Karplus The vast run for try9-opt1 is done. The best (most usable, longest) alignments are for 1nar and 1k87A ----------------------------------------------------------------- Mon Jul 12 12:00pm Jenny Draper Actually, try11-opt looks pretty decent! It seems to want to use "s4a" (153-156) rather than "s4" (125-129). It's that area of the barrel that's all torn apart... Tue Jul 13 00:14:16 PDT 2004 Kevin Karplus That's because the try10/try11 cost function TELLS it to use s4a. I've started a try12 run with the alignments from the try9 vast run. They may end up beating try9, though it seems doubtful, since it is starting from the wrong alignments. I'm going to kill it and try a different approach for initializing the try12 run---doing the TryAllAlign after reading the try9-opt1.pdb file. ----------------------------------------------------------------- Mon Jul 12 12:00pm Jenny Draper >That's because the try10/try11 cost function TELLS it to use s4a. ... which seems to be the right thing to do with it, is what I meant... ;) From learithe@soe.ucsc.edu Tue Jul 13 00:13:27 2004 MIME-Version: 1.0 Date: Tue, 13 Jul 2004 00:13:26 -0700 (PDT) From: Jenny Draper To: Kevin Karplus Subject: Re: T0208 status In-Reply-To: <200407130611.i6D6BGIX019817@cheep.cse.ucsc.edu> I'm not really sure what I should be doing at this point. Any suggestions, or should I just let you go at it? You seem to be getting waaaaay more done tonight than I've managed all week... -Jenny -------------------------------------------------- From karplus@soe.ucsc.edu Tue Jul 13 00:27:58 2004 Date: Tue, 13 Jul 2004 00:27:57 -0700 From: Kevin Karplus To: learithe@soe.ucsc.edu CC: karplus@soe.ucsc.edu In-reply-to: (message from Jenny Draper on Tue, 13 Jul 2004 00:13:26 -0700 (PDT)) Subject: Re: T0208 status Could you perhaps create a hand-tweaked model that is a better approximation to what try10 and try11 are trying to do than try11-opt1 managed to get? I'm not sure how you'd do that, but maybe like how you created the try8-try2-hand4.pdb file? I suppose you could try to find a model in which strand s4a exists, then superimpose and cut-and-paste to clean up the broken part of try11-opt1. S5 and S6 don't seem to be lining up the way I requested yet either. -------------------------------------------------- From learithe@soe.ucsc.edu Tue Jul 13 00:37:10 2004 MIME-Version: 1.0 Date: Tue, 13 Jul 2004 00:37:08 -0700 (PDT) From: Jenny Draper To: Kevin Karplus cc: learithe@soe.ucsc.edu Subject: Re: T0208 status In-Reply-To: <200407130727.i6D7RvxC020530@cheep.cse.ucsc.edu> Sure, but I'm not quite sure what you wanted try10/try11 to do? I'm assuming you were looking for this sheet from the README: s1 mKWGFRWYg s2 ALLGIES s3 iSLVCYs s4a kgFRLp s5 iKMGIHp s6 gITFCtg s7 NFVHFRNVK s8 eGVIRpd s1 mKWGFRWYg as opposed to this one: s1 mkwgFRWYg s2 alLGIEs s3 iSLVCYS s4 GSLSLLf s5 iKMGIHp s6 ANGITFCTg s7 NFVHF s8 egVIRPD s1 mkwgFRWYg What, specifically, are you looking for in try10/11 that you aren't getting? I want to be sure I'm trying the right stuff... -Jenny =============================================================== On Tue, 13 Jul 2004, Kevin Karplus wrote: > > Could you perhaps create a hand-tweaked model that is a better > approximation to what try10 and try11 are trying to do than try11-opt1 > managed to get? I'm not sure how you'd do that, but maybe like how > you created the try8-try2-hand4.pdb file? > > I suppose you could try to find a model in which strand s4a exists, > then superimpose and cut-and-paste to clean up the broken part of try11-opt1. > > S5 and S6 don't seem to be lining up the way I requested yet either. > ------------------------------------------------------------ From karplus@soe.ucsc.edu Tue Jul 13 00:59:58 2004 Date: Tue, 13 Jul 2004 00:59:56 -0700 From: Kevin Karplus To: learithe@soe.ucsc.edu CC: karplus@soe.ucsc.edu In-reply-to: (message from Jenny Draper on Tue, 13 Jul 2004 00:37:08 -0700 (PDT)) Subject: Re: T0208 status Other than the garbling of the strand alignment due to indentation, [Fixed in the copy saved in README] you are correct in what I'm trying to get. Following the barrel around in try11-opt2 I see Hbond s1-s2 W7 || I64 OK, though s1 is a bit twisted s2-s3 I64 || V102 OK s3-s4a V102 || F154 OK, but s4a is rather curled s4a-s5 R155 || M202 should be F154 || M202 s5-s6 I206 || G241 should be I206 || I242 s6-s7 C245 || V271 should be F244 || V271 s7-s8 V271 || I314 OK s8-s1 I314 || ? should be I314||W3 ----------------------------------------------------------------- Mon Jul 12 12:00pm Jenny Draper Got it. Working on it now. It'll take awhile; doing this stuff by hand is slow (but I'm getting faster ;) ------------------------------------------------------------ Tue Jul 13 01:04:15 PDT 2004 Kevin Karplus Now that I see try11-opt2, I'm tempted to try for yet another barrel structure: s1 mKWGFRWYg s2 ALLGIES s3 iSLVCYs s4a kgFRLp s5 IKMGIHp s6 gITFCtg s7 NFVHFRNVK s8 eGVIRpd s1 mKWGFRWYg SheetConstraint W3 Y8 A60 E65 hbond F5 SheetConstraint L62 S66 S100 Y104 hbond L62 SheetConstraint V102 Y104 F154 L156 hbond V102 SheetConstraint F154 L156 I202 M204 hbond F154 SheetConstraint M204 H207 I242 C245 hbond I206 SheetConstraint I242 C245 N269 H272 hbond I242 SheetConstraint V271 R274 G312 R315 hbond V271 SheetConstraint G312 R315 W3 R6 hbond G312 I'll put this in try13.costfcn. Tue Jul 13 01:23:29 PDT 2004 Kevin Karplus try13 started on cheep. ----------------------------------------------------------------- Mon Jul 12 2:00am Jenny Draper I've got a new superposition model: decoys/T208-try11-try4-try2-super.pdb which is a blend of: try11-opt1: 1-110 (s1,2,3) and 261-357 (s7,8) try4-opt2: 175-260 (s5,6) try2-opt2: 111-174 (s4a) I'll start playing with in in DeepView now... ------------------------------------------------------------ Tue Jul 13 02:24:57 PDT 2004 Kevin Karplus try13 crashed because T0208-try4-try2-super.pdb was not a legal pdb file. I renamed T0208-try4-try2-super.pdb to T0208-try4-try2-super.broken-pdb because "break here" that Jenny left on atom 1601 causes undertaker to die on it. Restarting try13. Mon Jul 12 3:30am Jenny Draper Damn, forgot to remove that. If you see something like that, just take it out... OK, I'm done with a hand-edited version of my try11-superpositon: T208-try11-try4-try2-super-hand.pdb Mostly, I just tweaked s4a so that it actually fit into the sheet. Let me know if it works, or if you want something fixed; if it's okay, I'd like to get some sleep... -------------------------------------------------- Tue Jul 13 04:08:04 PDT 2004 Kevin Karplus I'm submitting try10-opt2 and T208-try11-try4-try2-super-hand.pdb to vast. try10-opt2: VS59753 password T0208try10 T208-try11-try4-try2-super-hand.pdb: VS59755 password T0208super Tue Jul 13 07:39:18 PDT 2004 Kevin Karplus VAST claims I have the wrong password for VS59753, and hasn't finished running VS59755 after 3 1/2 hours! What a waste of time---time that I don't have to spare! I have submitted ReadConformPDB T0208.try13-opt2.pdb # best with try13.costfcn ReadConformPDB T0208.try12-opt2.pdb # best with try12.costfcn ReadConformPDB T0208.try3-opt2.repack-nonPC.pdb # Rosetta hates least ReadConformPDB T0208.try1-opt2.pdb InFilePrefix ReadConformPDB T0208.t2k.undertaker-align.pdb model 1 replacing previous submissions. I've also given VAST the try13-opt2 model VS59761 password T0208try13 Tue Jul 13 09:57:04 PDT 2004 Kevin Karplus Either the VAST queue is greatly overloaded, or their machines are wedged---I haven't been able to get anything for several hours. I was hoping to get an alignment corresponding to try13, but it seems unlikely now. Tue Jul 13 11:00:21 PDT 2004 Kevin Karplus The super-hand vast run has finished! Best alignments are to 6reqA (saved as 1reqA), 1de5A, 1a0dA, ... Of these only ReadFragmentAlignment 1a0dA/T0208-1a0dA-super.vast-a2m NOFILTER SCWRL produces a model with good sheets. I don't know that it is enough better to replace any of the top 4 models, but I will replace model 5. Sun Sep 19 12:56:22 PDT 2004 Kevin Karplus I put 1tz9A in the Makefile as the REAL_PDB entry, and evaluated whole-chain rmsd. Order of models is model5 (incomplete), model4, model1, model3, robetta1, model2, robetta7, robetta5, robetta6, robetta8, robetta10, robetta2, robetta3, robetta9, robetta4. Once again the fully automatic model is beating our hand-improved models! Repacking with rosetta does not seem to have helped. Our best complete model is actually try1-opt1! Wed Sep 22 11:22:41 PDT 2004 Kevin Karplus Using GDT score the order is try1-opt1-scwrl 36.7006% try1-opt1 36.6279% why do these differ? model4 36.1192 model1 27.7616 robetta2 27.3256 robetta6 robetta8 robetta3 robetta4 robetta1 model5 25.5814% model3 25.3634% robetta9 robetta10 robetta5 robetta7 model2 18.7500% The try13 costfcn was clearly better than the try12 costfcn (model1 better than model2) and the Rosetta energy function did not help pick out a good model (model3 does poorly). Our full auto was better than what we did by hand. Fri Sep 24 17:35:09 PDT 2004 Kevin Karplus Looking at the CDT curves and smooth_GDT scores, our model1 looks a little better than robetta, but model4 is a lot better: name length missing_atoms rmsd rmsd_ca GDT smooth_GDT T0208.try1-opt1.pdb.gz 357 0.0000 11.3728 10.8981 -34.7384 -32.9446 model4.ts-submitted 357 0.0000 11.4352 10.9385 -34.3750 -32.6938 full auto robetta-model2.pdb.gz 357 0.0000 17.8567 17.3048 -26.5262 -25.0347 model1.ts-submitted 357 0.0000 15.0086 14.3092 -25.5087 -24.9182 robetta-model6.pdb.gz 357 0.0000 17.0634 16.6102 -26.2355 -24.6041 robetta-model3.pdb.gz 357 0.0000 17.8595 17.3394 -25.5087 -24.2146 robetta-model8.pdb.gz 357 0.0000 17.4540 16.8290 -25.9448 -24.1575 robetta-model4.pdb.gz 357 0.0000 20.2278 19.7668 -25.5814 -23.9614 robetta-model1.pdb.gz 357 0.0000 15.4611 14.9681 -25.6541 -23.7383 model3.ts-submitted 357 0.0000 15.0696 14.4058 -24.2006 -23.1589 robetta-model9.pdb.gz 357 0.0000 18.0660 17.6352 -24.5640 -22.9509 robetta-model10.pdb.gz 357 0.0000 17.7151 17.2871 -23.9826 -22.8455 robetta-model5.pdb.gz 357 0.0000 16.3171 15.7432 -24.2006 -22.7143 model5.ts-submitted 357 593 9.8001 9.1653 -23.3285 -22.3144 robetta-model7.pdb.gz 357 0.0000 16.2602 16.0614 -23.0378 -21.8075 model2.ts-submitted 357 0.0000 15.8071 15.2020 -17.0785 -16.7105