Wed Jun 16 10:17:16 PDT 2004 T0206 DUE 16 Aug 2004 Wed Jun 16 10:43:22 PDT 2004 Kevin Karplus The searches are not done yet, but the t04 key residues list is showing 19 conserved glycines (23,26,29,32,35,38,41,44,47,50,53,56,59,62,65,68,71,74,80) from the GPT repeat. Probably because (as the T0206.doc.html file says) "The first 80 residues are collagen-like." This protein is a trimer, so predicting its structure with undertaker may be tricky. If we get a good model, we can superimpose it on a trimeric template and repack the sidechains with Rosetta. Wed Jun 16 13:04:30 PDT 2004 Kevin Karplus The t2k.best-scores all seem to be for the same family b.22.1.1, which is not hitting on the collagen-like part, but on the main body of the protein. This one may turn out to be a fairly easy comparative model (ha---none of the comparative models have been easy so far, so why should this one be?). The GPT repeat at the beginning may be modelable from fragments. The 1pk6 pdb file is a trimer, and may serve as a template for building a trimeric structure. The N-terminal tails should look something like the trimeric collagen structure in 1cag, though the HYP residues there may make different backbone bends than the THR residues in T0206. Wed Jun 16 14:42:06 PDT 2004 Kevin Karplus The try1 model looks pretty good in the globular part using the t2k colorings, but the t04 secondary structure predictions look poor. I'll have to make some sort of option in the Makefiles to specify whether the t2k or the t04 rasmol files get the softlinks (currently it defaults to t04, which is often not good). For the collagen repeat part, maybe I should add some P-P distance constraints for adjacent and +2 repeats. It looks (from 1cag) that the Px.CG Px+3.CG distance is about 9.5 (but sometimes as big as 10.5) Px.CG-Px+6.CG distance is about 17.1. The Gx.CA atoms are in almost a straight line, with spacing 8.6 +- 0.2. It may be difficult to get both the collagen-like part and the globular part (both naturally trimeric) to line up properly for forming a trimer. I think we can do superposition to get a trimer based on the natural trimerization of each part separately (using undertaker scripts similar to the ones I used in pce/protein-predict/ach/ach7_short_1based/). The we could look at constraints needed between the two parts to get the collagen-like part departing at a suitable angle to make the trimerizations compatible. Thu Jun 17 06:38:58 PDT 2004 Kevin Karplus Despite adding lots of constraints, the collagen-like part is still crumpling up. Some of the sandwich is also getting messed up, perhaps because of the relative weakening of the strand constraints, with all the new constraints added. Let's try again with some long-distance constraints added for the collagen-like part, and with the 42 shorter collagen-like constraints turned down by a factor of 50, to make them more comparable with the strand constraints. Thu Jun 17 21:06:15 PDT 2004 Kevin Karplus I'll definitely have to split this one 1-90 81-end. It is too hard to do the extended collagen structure and the main body simultaneously. Putting the pieces back together will be tough---I can trimerize each separately, but then the superposition of the two trimers will be difficult to arrange. Tue Jul 27 10:30pm Jenny Draper Domain2 (the non-collagen, 81-end domain) has hits mainly to SCOP domains b.1.18.8 and b.22.1.1 -- for b.22.1.1, it matches the trimeric structure 1kxgA (1pk6, mentioned above, is also domain b.22.1.1, but doesn't seem to match just the second domain as well as 1kxgA). -- thus, 1kxg might be a good template for the trimer superpositioning for domain2. Try1 for domain2 looks not-so-good; it's trying to coil up all the strands. I'm running try2 for this domain from the alignments (including the complete "best match" alignments), with the same basic costfunction as (whole)try3 -- just strand and helix constraints. I'm also put the domain2 conformations generated by (whole)try3, and robetta model 2 (their model 1 incorporates the helix into the sandwich, but model 2 looks good) into the 81-end/decoys directory, and in it's read-pdb.under file. I'll start try3 from these, if try2 doesn't start working out. Wed Jul 28 12:00am Jenny Draper I was trying to use the "read-alignments.under" files in the Top-Hits directories for domain2 -- but they don't exist. After editing the ID's of the "Manual_Top_Hits" in the 81-end/Makefile, then "make read_alignments" generated "read-alignments-noscwrl.under" and "read-alignments-scwrl.under" in each directory -- but no plain "read-alignments.under". What's going on here? ... I'm going to edit try2.under to read the "scwrl" versions... Sun Aug 1 4:20pm Jenny Draper Try2 is hideous. I'm going to set up a try3, running from the previous structures, and still using the try2 costfunction. Mon Aug 2 5:20pm Jenny Draper Try3 looks exactly like try2; apparantly it started from try2. Damn. I'll try again, this time with some sheet constraints picked from Robetta-2 and whole-try3. Wed Aug 4 7:20pm Jenny Draper I really like the structure generated by the alignments to domain2 for the whole-chain. They all agree nicely and match the secondary structure well. How do I snag this core structure to start from? The alignments for 81-end are kinda all over the map, beyond a core that is similar to that in the whole structure. Wed Aug 4 11:00pm Jenny Draper I created a new domain2 structure: 81-end/decoys/wholestruct_alignModel1_try1_superpose.pdb by superimposing the beginning (81-85) and end (203-220) of domain2 from whole-try1-opt2 onto the structure from alignment model 1 (from whole T0206.t2k.undertaker-align.pdb.gz) -- using, of course, DeepView's magicFit to do the superpositioning, because it's fast and I can see and modify what it's doing. Th Aug 5 1:00am Jenny Draper Using the try4 costfunction, which includes some sheet constraints from the whole-structure best alignment model, my superpositioned structure scores the best, so I'm running try4 off of all current models (read-pdb.under), figuring that it'll start from there. Th Aug 5 12:00pm Jenny Draper Try4 looks pretty good! The sheets still need some polishing; a few strands have been coiled up into helices, but it's a huge leap forward. The region 81-96 forms a strand in the sheet, but is predicted to be coil/turn. It's possible that this should lift out of the sheet, and s1 should take it's place. We should definitely try this at some point. The region 164-186, where there is little signal, is a long sheet in the template/alignment model 1; but it's coiled up and pulled out here. Th Aug 5 12:00pm Jenny Draper I've done some tweaking of the sheet constraints in try5; we'll see how it goes... Fri Aug 6 12:00am Jenny Draper Try5 looks better, except I definitely need to work out the placement for strands s2, s7, and s8. Particularly s7, which I've got trying to go in two opposing directions at once... The loop region between s2 & s3 needs to be straighted out; it's coiling up into a helix. It looks like s2 should pair with s3, with s1 remaining on the other side of the sandwich. Hmmm. could s7 pair up with s5? There were some hints of that earlier... Fri Aug 6 1:00am Jenny Draper Starting try6, changing: * halved weight of existing strands * made s1-s0 sheet bonus * made s7-s5 pairing * killed s1-s2 pairing * upped weight of s2-s3 pairing Sat Aug 7 15:31:07 PDT 2004 Kevin Karplus It seems that someone did tell Jenny about "make extra_alignments", since all the alignments were there, but I did a 'make all-align.*' also. Perhaps she needs it in one of the subdirectories? All her notes above seem to be for the subdirectories, and not for the whole protein, as the most recent thing there is the try3-opt2 that I did way back in June. It is probably time for Jenny to create a trimer of each domain and paste them together. We need to submit both the monomer and the trimer for this target (see ../mail), so we'll need some extra time for the submission. Mon Aug 9 15:34:23 PDT 2004 Kevin Karplus I'm redoing the "make" on T0206, 1-90/, and 81-end/ since the previously done one was long enough ago that several things were not working at that time. Tue Aug 10 4:00pm Jenny Draper I've superimposed domain2 try6 on 1jh5 (best alignment for domain2) 1kxg (second-best), and 1pk6 (the one Kevin liked for the whole-chain, which doesn't score well for domain2 only); these are all scop domain b.22.1.1). Trimers from 1jh5 & 1kxg are exactly the same, but the model from 1pk6 is a mirror image, and slightly different in internal structure. All overlap badly in a few places. From learithe@soe.ucsc.edu Thu Aug 12 11:12:33 2004 MIME-Version: 1.0 Date: Thu, 12 Aug 2004 11:12:30 -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: help with T0206 In-Reply-To: <200408082132.i78LWZw7000893@cheep.cse.ucsc.edu> Hello everyone, I haven't been able to work on T0206 for the last two days, and I can't work today, as always.... so if anyone could take a look at T0206 for me, I would really appreciate it. I've set up a few trimers, all of which have bad backbone clashes. I think my current model for domain 2 (81-end) should look more like the best alignment models; my sandwich has deviated quite a bit from the alignment's structure. -Jenny Thu Aug 12 17:48:17 PDT 2004 Sol Katzman Created a chimera from domain1 try1 and domain2 try6 using DeepView: (the 'dv2' means it was my second attempt. forget about dv1) decoys/T0206.d1.try1.d2.try6.dv2.pdb I just eyeballed the approximate attachment angle as I had only loaded the d2.try6 monomer. In future attempts, I will load the d2 trimer and maybe a d1 trimer (when I make one) to glue them together better. Thu Aug 12 22:04:52 PDT 2004 Sol Katzman In creating 81-end try7, I noticed that cost function weights for hbond_geom_beta* were based on the old undertaker. Increased them to what they should be for current version of undertaker. Looking at trimer-opt6 we need to move S166-Q177 over to be closer to T142-L151 to avoid knotting with other monomer. This can probably be done with SheetConstraint: SheetConstraint I174 A178 L148 T144 8 # s7a ^v s4b I tweaked the strand constraints to take into account some of the t04.str2 predictions. And increased the weights on the SheetConstraints which were very low, considering that they get dispersed among individual residues and atoms. Date: Thu, 12 Aug 2004 21:03:21 -0700 To: Kevin Karplus From: Sol Katzman Subject: need 1cagA directory for T0206 1-90 alignment Dear Kevin, Can you run whatever is necessary to get 1cagA/info so I can use it as a trimer template for T0206/1-90 /Sol. ---------------------------------------------------------------- Error: Couldn't open file /projects/compbio/experiments/models.97/pdb/1c/1cagA/info/1cagA.stride-mixed.seq: No such file or directory Date: Fri, 13 Aug 2004 08:13:07 -0700 From: Kevin Karplus To: katzman@ieee.org Subject: Re: need 1cagA directory for T0206 1-90 alignment 1cagA, which is identical in sequence to the better resolution 1cgdA is a short peptide, but it may be useful for trimerization. Collagen-like peptides in dunbrack's pdbaa set include >1ei8A 29 XRAY 2.00 0.251 0.289 COLLAGEN-LIKE PEPTIDE (PRO-HYP-GLY)4-PG-(PRO-HYP-GLY)5 [NA] || 1EI8B 1EI8C 1EI8D 1EI8E 1EI8F >1bkvA 30 XRAY 2.00 0.228 0.277 T3-785 [NA] || 1BKVB 1BKVC >1cgdA 30 XRAY 1.85 0.172 NA COLLAGEN-LIKE PEPTIDE [NA] || 1CAGA 1CAGB 1CAGC 1CGDB 1CGDC >1k6fA 30 XRAY 1.30 0.226 0.297 collagen triple helix [NA] || 1K6FB 1K6FC 1K6FD 1K6FE 1K6FF >1nayA 56 XRAY 2.60 0.234 0.296 Collagen-like peptide [NA] || 1NAYB 1NAYC >1qsuA 30 XRAY 1.75 0.180 0.227 PROTEIN ((PRO-HYP-GLY)4- GLU-LYS-GLY(PRO-HYP-GLY)5) [NA] || 1QSUB 1QSUC I've done serial-make basic-setup for the whole list, but 1bkvA 1qsuA 1k6fA 1ei8A are already in the template library. Try 1k6fA as a template for the trimerization. ___________________________________________________________________________ Fri Aug 13 08:57:16 PDT 2004 Sol Katzman Thanks, Kevin. For the 1-90 domain, I did 'make extra_alignments' after setting this in the 1-90 Makefile: MANUAL_TOP_HITS:= 1ei8A 1bkvA 1cgdA 1k6fA 1nayA 1qsuA The 81-end try7 run may finish soon. It has been running over 12 hours. For 81-end try8, I will reduce super_iter and super_num_gen so we can get another result before this target expires. Also, I have made some more educated guesses about the sheets. The only thing I find odd is the parallel s0 || s4a, but in try7-opt1 it looks quite feasible. These are the strands (defined in a rasmol script pointed to by 'strands') define s0 91-93 define s1 109-111 define s2 118-121 define s3 125-128 define s4a 134-140 define s4b 144-148 define s5 151-155 define s6 158-162 define s7a 174-178 define s7b 181-185 define s8 189-193 define s9 199-202 And I hope to get them to form 2 sheets of a sandwich as follows, with the indicated bonding patterns: SHEET1: s1^ s0v s4av s8^ s7bv s7a^ s1 111 110 109 | | s0 88 89 90 91 92 93 / \ / \ / \ s4a 134 135 136 137 138 139 140 | | s8 193 192 191 190 189 | | s7b 181 182 183 184 185 | | | s7a 178 177 176 175 174 SHEET2: s2^ s3v s9^ s5v s6^ s2 118 119 120 121 | | s3 128 127 126 125 | | s9 198 199 200 201 202 | | s5 155 154 153 152 151 | | | s6 158 159 160 161 162 For 81-end try8, other minor changes in the try8.under file include the output of Sheets and Helices. Also, I am going to read in the t04 fragments (t04.many.frag). 81-end try7 finally completed. It still has the knots with the other monomer when I made the trimer-1jh5 model. One possibility is that we are in a local minimum, so in parallel with try8, I will launch try9 with the same cost function but using TryAllAlign, rather than read-pdb.under. Also, I believe the short helix at G164-S169 between strands s6 and s7a is in the wrong position. Because of this, the strands s7a, s7b, s8 cannot form in the directions that I want them in SHEET1, so s7a, s7b are "upside down", and s8 does not form properly. I will add some scaffolding constraints on this helix. Maybe S169 and S146 should be close. That will be try10 for 81-end. Meanwhile I completed making the trimer for 1-90, using the 1k6f template. It does not look terrible but it is knotted up among the monomers at the N-terminus where we have some sort of paperclip type loop in our try1 model at M1-T34. After a discussion with Kevin about 1-90, the plan is to lose the paperclip. We will simply create a series of short (20 residue) trimer aligments to compose the full trimer. We will create a chimeric collagen-like trimer by pasting the several of these together in Deepview. Next we will paste the 1-90 (really 1-80) trimer to the 81-end trimer. That will be our TRIMER submission. From that we will select one chain as our best MONOMER submission. Regarding 81-end, Kevin also pointed out that in the alignments, the region G164-S169 is often a strand in a sheet, not a helix at all. Fri Aug 13 16:37:18 PDT 2004 Sol Katzman For 81-end try9 and try10 have completed and look terrible. Not many of the desired sheets have formed, and the trimer model has much more conflict than try7. Try8 has not yet completed, but it does not look like I will be able to improve on try7. Sat Aug 14 12:30am Jenny Draper I'm not sure from this readme where work on domain1 is at, but I suggest you use 1bkvA as the template, as it has the (PxG) sequence that T0206 has (PTG); 1k6fA has the sequence (PPG)... ? Sat Aug 14 13:31:19 PDT 2004 Sol Katzman I tried to take Jenny's suggestion about using 1bkv as the trimer template for 1-90 but got (see T0206/1-90/make-trimer-1bkv-try1.begin.log) Error: PrintMultimerPDB requested a 3-mer, but the alignment library has only 0 alignments I think this is because of CHAIN_BREAK_BEFORE at many of the glycines in 1bkv. Anyway, 1bkv uses PHG, where H (I now know, thanks to Kevin) is hydroxyProline, so this is not that much different from PPG. Using 1k6f as the trimerization template for 1-90, I made several versions of the alignment (since 1k6f is much shorter than 90 residues) to generate several trimers: trimer-align61-80-1k6f-try1.pdb trimer-align41-60-1k6f-try1.pdb trimer-align21-40-1k6f-try1.pdb trimer-align01-20-1k6f-try1.pdb For these, I measured interchain distance of CA atoms at key residues 81,61,41,21. I also measured the same distances at residue 81 for the 81-end domain trimerizations. Here are the numbers (AB, BC, CA refer to chain-pairs of interchain measurements) chain-pair -> AB BC CA --- ---- ---- 1-90 trimers: ------------- trimer-align61-80-1k6f-try1.pdb (same as align41-60) trimer-align41-60-1k6f-try1.pdb L81.CA 4.8 4.6 7.3 T61.CA 5.3 5.4 8.0 G41.CA 27.6 27.2 33.2 P21.CA 59 59 72 trimer-align21-40-1k6f-try1.pdb L81.CA 27 28 34 T61.CA 4 4 6.7 G41.CA 6.4 6.4 9.1 P21.CA 16.6 16.6 20.9 trimer-align01-20-1k6f-try1.pdb L81.CA 76 79 100 T61.CA 65 70 87 G41.CA 30 32 41 P21.CA 8.2 8.5 11 81-end trimers: ------------- trimer-1jh5-try6.pdb L81.CA 25.6 25.7 25.7 trimer-1jh5-try7-opt2.pdb L81.CA 26.1 26.1 26.1 Based on these numbers, it should be pretty easy to make a chimer by attaching the align21-40 trimer of 1-90 to the trimer of 81-end. Furthermore, it does not seem useful (and seems difficult) to make a chimer by attaching the subalignments of 1-90. Finally, the align21-40 trimer only spreads out a little bit at its N-terminus, which in fact accomodates the (possibly bogus) paper clip motif from its first 20 (non-collagen-like-sequence) residues: MAFDPNLVGPTLPPIPPFTL That is, using align21-40, there are no major conflicts between the 1-90 monomers. With regards to the 81-end attempts, try8 did not improve matters, failing (like try9 and try10) to form any more desired sheets. ============================================================ Date: Fri, 13 Aug 2004 11:20:21 -0700 From: Kevin Karplus To: katzman@ieee.org Subject: Re: need 1cagA directory for T0206 1-90 alignment A quick google search of HYP PRO amino acid reveals that HYP is hydroxy-proline, which has an OH instead of one of the Hs on the CG atom. Collagen has 4-hydroxyproline and 3-hydroxyproline. I don't know the numbering system, so I don't know which carbon in C4 and which C3. hydroxy-Proline (Hyp) ? Repeating triplet: -(Gly-Pro ... www.mmb.usyd.edu.au/BCHM2611/jmg/jmg-lecture-06.pdf Date: Sat, 14 Aug 2004 13:08:21 -0700 To: Kevin Karplus From: Sol Katzman Subject: collagen 3-helix Dear Kevin, The pointer to the lecture notes you sent me about collagen was good. On page 15 of the presentation, they say collagen helix is LEFT-HANDED, but the pictures in the subsequent pages all look right-handed to me. And our superposition of T0206/1-90 are right-handed. ============================================================ Sat Aug 14 15:23:49 PDT 2004 Kevin Karplus I agree, 1k6f is definitely showing Z-twist, which is right-handed. ------------------------------------------------------------------- Sat Aug 14 16:07:38 PDT 2004 Sol Katzman I completed the construction of the chimeric trimer-trimer using DeepView. It was a little tricky to get the chains contiguous in the pdb file. I will add notes to the main casp6/README to explain how. The outputs are here: full trimer: T0206/T0206.3mer.ABC.d1.try1.d2.try6.dv3.pdb individual chains (monomers) of the full trimer T0206/decoys/T0206.3mer.C.d1.try1.d2.try6.dv3.pdb T0206/decoys/T0206.3mer.B.d1.try1.d2.try6.dv3.pdb T0206/decoys/T0206.3mer.A.d1.try1.d2.try6.dv3.pdb It looks like the chainA monomer has too much overlap between the domains. I think we should submit chainB or chainC. Sat Aug 14 17:36:16 PDT 2004 Kevin Karplus Of the three monomers extracted from the trimer, the 3mer.c.d1.try1.d2.... one scores best with an unconstrained cost fcn. Of the 81-end models, the best-scoring with an unconstrained cost fcn is try7-opt2. Is that what Sol used for the trimer? or did he use try6? Rosetta likes 81-end try1-opt2.repack-nonPC best. I'll submit T0206.3mer.C.d1.try1.d2.try6.dv3.pdb # best whole monomer 81-end/try7-opt2 # best-scoring globular domain 81-end/try1-opt2.repack-nonPC # best Rosetta energy for globular domain try1-opt2 # full auto T0206-1gr3A-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 # best template It looks to me like there is a bad strand pairing in the 81-end/try7-opt2 model (and hence the trimer) around P186 and L189. If you look at the superposition in best_models.pdb.gz, you can see that try1-opt2 does better in this region that 81-end/try7-opt2. I think that more damage than improvement was made in the subdomain optimization. I split off the 81-end part of try3-opt2 and put it in 81-end/decoys/from-whole-try3-opt2.pdb It scores almost as well as some of the more highly optimized I wonder if it is worthwhile to try trimerizing it? (or chopping of the end of try2-opt2 similarly). Hmm, maybe not. I tried superimposing the globular part of try3-opt2 on the whole monomer model and there are a lot of discrepancies. I'll submit what we have now, and ask Sol if he wants to do more. Maybe I should submit the globular part of try1-opt2 on the whole monomer, and splice it into the whole monomer. They superimpose well, and the try1-opt2 looks better, at least after about D124. I made a chimera: decoys/try1-whole-chimera.pdb which crosses over at D124 from T0206.3mer.C.d1.try1.d2.try6.dv3 to try1-opt2, but this chimera makes it fairly clear that there is a missing strand between F90-G93 and T137-A140. I wonder where that is supposed to come from, or whether we just have really bad misalignment. ----------------------------------------------------------- Date: Sun, 15 Aug 2004 07:35:54 -0700 To: Kevin Karplus From: Sol Katzman Subject: Re: T0206 submitted, but ... Dear Kevin, Yes, that does look like it has some of the sheets I was trying to make. If we had another few days...but I better look at T0254 and T0242 today. /Sol. At 07:24 PM 8/14/2004 -0700, you wrote: >I submitted T0206, but I'm not very happy with the globular part. >I think we have a better globular part from the fully automatic prediction. >I've put it in T0206/81-end/decoys/from-whole-try1-opt2.pdb.gz > >Take a look at it (particularly in the superposition >T0206/globular-on-whole-model.pdb) and see if you agree that it is a >a better model, at least after D124. Date: Sun, 15 Aug 2004 08:44:15 -0700 From: Kevin Karplus Subject: Re: T0206 submitted, but ... The big questions are should I submit T0206/decoys/try1-whole-chimera.pdb in place of our current model 1? should I re-optimize 81-end for that chimera and submit is as model 2? There is just enough time for me to do this. Date: Sun, 15 Aug 2004 09:54:26 -0700 From: Sol Katzman Subject: Re: T0206 submitted, but ... I would say yes to both if you have time. /Sol. ============================================================ Sun Aug 15 13:45:51 PDT 2004 Kevin Karplus I reoptimized the chimeric globular domain, and have a new best unconstrained score for 81-end/decoys/T0206.try11-opt2.pdb It has one awkward break before S187, but is otherwise fairly good It is possible that the C-terminal helix should really be a strand filling in the missing strand on the beta sheet, but forcing that would be too difficult in the time remaining. Sun Aug 15 14:37:07 PDT 2004 Kevin Karplus I pasted the new globular part into our best previous monomer, then trimerized it. The collagen-like parts diverged rather badly. I superimposed the trimer on the previous trimer, did a cut and paste of the best matching collagen-like domain, and got a new monomer. Trimerizing the new monomer still doesn't quite give us the collagen-like part, but it is fairly close. I've submitted the new decoys/whole-try11-chimera1.pdb as model 1 and the new 81-end/decoys/T0206.try11-opt2.pdb as model 2. If Sol can create a new trimer with the collagens closing, then I might pull out the improved monomer. From karplus@soe.ucsc.edu Sun Aug 15 14:42:01 2004 Date: Sun, 15 Aug 2004 14:42:00 -0700 From: Kevin Karplus To: sol@soe.ucsc.edu CC: karplus@soe.ucsc.edu Subject: fix up T0206 trimer Could you fix up the trimer for T0206? I have a new best monomer decoys/whole-try11-chimera1.pdb, but trimerizing it doesn't quite get the three collagen parts to superpose properly. The trimer is trimer-1jh5-whole-try11.pdb. I superimposed it on the old trimer (trimers-superposed2.pdb), but the superposition lined up one of the collagen parts, rather than centering the collagens in the middle. I think it would be a fairly easy matter to reposition the 1-80 part of the old trimer on trimer-1jh5-whole-try11.pdb then cut and paste to make a new trimer, but I don't have DeepView here to work with. Could you do that repositioning? ============================================================ Sun Aug 15 21:26:05 PDT 2004 Sol Katzman Looking at Kevin's trimer-1jh5-whole-try11, the interchain distances between L81.CA is listed below, along with the previously used collagen trimer. These should be quite compatible so I will merge them in DeepView. chain-pair -> AB BC CA --- ---- ---- 81-end trimers: ------------- trimer-1jh5-whole-try11.pdb L81.CA 32.8 32.9 32.9 1-90 trimers: ------------- trimer-align21-40-1k6f-try1.pdb L81.CA 27 28 34 T61.CA 4 4 6.7 G41.CA 6.4 6.4 9.1 P21.CA 16.6 16.6 20.9 From karplus@soe.ucsc.edu Sun Aug 15 21:35:25 2004 Date: Sun, 15 Aug 2004 21:35:23 -0700 From: Kevin Karplus To: sol@soe.ucsc.edu CC: karplus@soe.ucsc.edu Subject: T0206 trimer When merging the T0206 trimers, if you start with the trimerized whole chain, it should be easier to center the collagen, since you have the tunnel of three not-quite-meeting collagen parts to show you where the axis needs to be. ============================================================ Sun Aug 15 22:18:12 PDT 2004 Sol Katzman It is done. The full 3-chain result is in T0206/T0206.3mer.ABC.whole.try11.pdb From karplus@soe.ucsc.edu Sun Aug 15 22:46:13 2004 Date: Sun, 15 Aug 2004 22:46:13 -0700 From: Kevin Karplus To: katzman@ieee.org CC: karplus@soe.ucsc.edu In-reply-to: <5.1.1.6.1.20040815221433.021c9870@pop.sbcglobal.yahoo.com> (message from Sol Katzman on Sun, 15 Aug 2004 22:16:02 -0700) Subject: Re: T0206 trimer The trimer looks pretty good. I'll try submitting it. Thu Nov 18 22:20:12 PST 2004 Martina Koeva Based on the smooth gdt scores: best sam-t04 31.8853 (this was our align1) best submit 31.8819 (it looks like also align1?) model1 25.4751 auto 24.7654 align 31.8853 robetta best 46.0309 robetta1 46.0309 Our top alignment scored the best on this target.