Mon Jun 28 09:40:44 PDT 2004 T0217 Due 12 aug Mon Jun 28 11:57:47 PDT 2004 Kevin Karplus Looks like a fold-recognition target, possibly a TIM-barrel. Possibly a larger barrel or partial barrel (1oltA has 12 strands forming a half barrel). Mon Jun 28 13:40:50 PDT 2004 Kevin Karplus It looks like there are probably 8 strands, so a TIM-barrel is definitely possible, even if it is not one of the top 4 hits. This is rather hard for fold recognition, with little agreement among the top hits. Wed July21 11:54 2004 Bret Barnes So, I'm going to try to form a TIM-barrel for try2, but I'm also going to see what VAST or DALI gives me for structural alignments to try1. If I get a good hit to some to a barrel or partial barrel I'll do some trys for that. This will probably take me until tomorrow (Thurs July 22) to get running since I havn't really worked with TIM-barrels or VAST or DALI. For try2 (attempt to form a TIM barrel) I increased the strand constraint for strand A188 - F192 becasuse it does not appear in try1 and should serve as the 8th strand in the TIM barrel. I also added some constraints to put A188-F192 inbetween strand 164-168 and strand 234-238 (which fromed a sheet in try1). I made a new sheet constraint for strand 110-113 and strand 141-144 to close the gap in try1. Tue Jul 27 11:54:32 PDT 2004 Martina Koeva Hi Bret, T0217 came up yesterday in our CASP meeting and I just looked at it a bit ago. Usually when Kevin runs VAST on a target, he puts in the JobID and password for it in the README file for later reference or in case somebody else decides to look at it. I was wondering if you could put that in the README file. Also, did you have any success with DALI? I tried using it once for T0212, but that did not seem to work very well, as I was not able to find a way to transform the output from DALI into a format readble to undertaker. What was your experience? Martina Fri Aug 6 09:20:47 PDT 2004 VAST Search ID: VS60548 Password: T0217try2 Sat Aug 7 10:27:44 PDT 2004 Kevin Karplus I see no evidence that Bret has done anything on this protein since 21 July, so I am taking him off as lead. Temporarily, I've assigned myself to this protein, but I'll hand it over to someone else on Monday (possibly Martina). Bret started a try3.costfcn, but never used it. I've renamed it bbarnes-try3.costfcn, so that I can create a try3.costfcn and not have to skip a number. Martina submitted try2 to vast yesterday, and got some hits: 1k77A and 1fq0A=1euaA for the whole chain, and several for the 3 "subdomains" identified by VAST. The first domain (1-50) is too short to have decent matches, and the little matches that are found are very gappy. The second domain (50-150) has a few matches, but they are pretty gappy. One of the best is for 1l5yB=1l5zA. VAST did not find any matches for the 3rd domain. I've picked up the alignments for 1k77A, 1euaA, and 1l5zA, and added them to the MANUAL_TOP_HITS, along with many of the top t04 hits, which overlap only partially with the t2k hits. I'll make all the alignments for the top hits and update all-align.a2m.gz, so that we can try doing another prediction run with more starting alignments. I created an unconstrained.costfcn. With it, try2-opt2 scores the best. try2-opt2 forms a fairly decent parallel beta sandwich surrounded by helices, with conserved polars clustering around the C-terminal ends of the strands. The rr predictions look plausible with this model, so I'll toss them into the try3.costfcn. I'll also redo the initial make, so that the t04.many.frag file is created, and include it in the try3.under file. I don't see a big difference between the t2k and t04 conservation patterns, so it is not clear which multiple alignment I should prefer. Sat Aug 7 13:26:52 PDT 2004 Kevin Karplus oops, the remake did not create T0217.t04.many.frag.gz. I've fixed Make.main and will create that file. Then try4 will have to be run to take advantage of it. Sun Aug 8 03:55:36 PDT 2004 Martina Koeva This will be just a quick note (since it is quite late) about the two new runs. I have started try4 and try5 simultaneously on two different machines. Try5 is the same as try3, except for: 1.) T0217.t04.many.frag.gz is in place and included in the try5.under file. 2.) I have increased the phobic_fit weight from 2 to 4, as well as the break weight from 40-50. Try4 has the above 2 changes, as well as additional sheet constraints that I thought I liked. Try3 is done. I looked at the try3-opt2 and I actually quite liked the barrel that was formed (it was quite pretty), even though the structure really did not match our secondary structure predictions. It also scored worse than try2-opt2 with the try3 cost function. I will put in more detailed comments on the sheet constraints, templates and literature in the morning. Sun Aug 8 06:59:58 PDT 2004 Kevin Karplus The barrel in try3-opt2 looks pretty badly mis-aligned---the secondary structure predictions don't match at all. The try4 run looks like it was killed before producing any output other than the log file. The try5-opt1 barrel also seems to have misalignment. Perhaps the strand constraints need a higher weight, so that TryAllAlign can pick out a better alignment. Although try2-opt2 is not as pretty a barrel, it is much more faithful to the predicted secondary structure. It also scores best with try2.costfcn, try3.costfcn, try4.costfcn, but not try5.costfcn (which prefers try3-opt2). We may want to pick up sheet constraints from try2-opt2.sheets, perhaps supplementing them with some almost-formed strand pairs, and use them for the next try. I'll leave this for Martina to work on. Sun Aug 8 14:31:27 PDT 2004 Martina Koeva I restarted try4, since no files had really been created yet and I did not think there was a need for renumbering the try. Try5 was supposed to follow try3 in that it did not have any specific sheet constraints, but only the helix, strand and rr constraints. In try4, I decided to keep some of the original sheet constraints from try1-opt2.sheets (s8 || s1 || s2 || s3 || s4): SheetConstraint V2 S7 E233 F238 hbond Y3 7 SheetConstraint G5 S7 I40 E42 hbond T6 7 SheetConstraint R39 L43 F67 V71 hbond V41 7 SheetConstraint T70 T73 M110 A113 hbond V71 7 The other 4 constraints came from my hand-aligning the rest of the strands and looking for the possible strand pairings. Here is what I came up with: TLAQ | LAVEF | TFVVV | YAYFR | ETYV It might not really make any sense, but I thought it was worth a try. Here are the last 4 sheet constraints (s4 || s5 || s6 || s7 || s8): SheetConstraint L112 Q114 L140 V142 hbond A113 5 SheetConstraint A141 R145 T164 V168 hbond V142 7 SheetConstraint F165 V168 Y187 F190 hbond V166 5 SheetConstraint A188 F191 E233 V236 hbond Y189 5 Sun Aug 8 15:13:20 PDT 2004 Kevin Karplus try4-opt1 is looking pretty good. The barrel is almost formed and it is consistent with the secondary structure predictions. I think that the strand that has conserved R191 is oriented the wrong way---the R191 should point into the barrel. Similarly for the strand with E143 and R145 It would probably be worthwhile to make a "strands" script to name the strands, and to show the intended parallel alignment here in the README file, so that we can check that the conserved polars are all on the same side (inside) of the barrel and that the slip is 8 going around the 8 strands of the barrel. Sun Aug 8 19:24 2004 Bret Barnes So, my new sheet constraints for try6 are very similar to try4, but I also added some constraints to put R191, E143 and R145 all on the right side of the sheet. I also, ran a vast alginment for try5 (before try4 finished), its taking a while to finish, but once it does I'll add it to the alignment (that is if it finds different targets). ID: VS60576 Password: bb7509bb Running try6 on woof. Problem: I can't use the Template.atoms file, I belive this is because I don't have permission to write to that file. I'll ask Martina if she can fix that. Mon Aug 9 15:41:16 PDT 2004 Martina Koeva This alignment of the strands should be able to flip strands 5 and 7, which have the conserved polar residues (in particular, E143 and R191). I have preserved the alignment of the first 4 strands and the last 2 strands and have changed the alignment of s5, s6 and s7. s1: VYVGTS | s2: RIVEL | s3: FYFTV | s4: MTLA | s5: LAVEF | s6: TFVVV | s7: YAYFR | s8: ETYV I have also defined a "strands" rasmol script for this target. Mon Aug 9 16:38:49 PDT 2004 Kevin Karplus I like this arrangement better: oioioioioioioioi 1: MVYVGTS 39: rIVELNF 67: FYFTVK 110: MTLAQF 139: eLAVEFR 164: TFVVVDe 187: YAYFRFH 233: ETYVFFN 1: MVYVGTS This implies the following sheet constraints, if the hbonding is on the outside residues of the earlier strand and the inside resides of the later strand in each pair (as in try4-opt2): SheetConstraint Y3 S7 I40 N44 hbond V4 SheetConstraint V41 F45 F67 V71 hbond V41 SheetConstraint Y68 K72 M110 Q114 hbond F69 SheetConstraint T111 F115 L140 F144 hbond T111 SheetConstraint L140 R145 T164 D169 hbond V142 SheetConstraint F165 D169 Y187 R191 hbond V166 SheetConstraint Y187 H193 E233 N239 hbond A188 SheetConstraint Y235 N239 M1 G5 hbond V236 Mon Aug 9 17:23:09 PDT 2004 I have started try7 with the new sheet constraints for the barrel. Mon Aug 9 18:24 2004 Bret Barnes After reviewing the alignment that Martina came up with I moved a couple of strands slightly and added a little of length to some strands. Bellow is an alignment of how I think the strands might go. This seemed like the best alignment to try to put all the polar residues on one side and the non-polars on the other. Polarity NP P NP P NP P NP P NP P NP P NP P NP S1: 2 -8 V Y V G T S G S2: 39 - 43 R I V E L S3: 67 - 72 F Y F T V K S4: 109 - 113 K M T L A S5: 139 - 144 E L A V E F S6: 163 - 168 I T F V V V S7: 187 - 191 Y A Y F R S8: 233 - 238 E T Y V F F S1: 2 -8 V Y V G T S G Constraints used in try6. SheetConstraint (T0217)V4 (T0217)G8 (T0217)R39 (T0217)L43 hbond (T0217)T6 SheetConstraint (T0217)R39 (T0217)L43 (T0217)F67 (T0217)V71 hbond (T0217)E42 SheetConstraint (T0217)F69 (T0217)K72 (T0217)K109 (T0217)L112 hbond (T0217)V71 SheetConstraint (T0217)M110 (T0217)A113 (T0217)E139 (T0217)V142 hbond (T0217)L112 SheetConstraint (T0217)L140 (T0217)F144 (T0217)I163 (T0217)V167 hbond (T0217)V142 SheetConstraint (T0217)T164 (T0217)V168 (T0217)Y187 (T0217)R191 hbond (T0217)V166 SheetConstraint (T0217)Y187 (T0217)R191 (T0217)E233 (T0217)237F hbond (T0217)F190 SheetConstraint (T0217)T234 (T0217)238F (T0217)V2 (T0217)T6 hbond (T0217)F237 If you look at the sequences after the C-terminus of each strand you will find several charged residues that could work together to make an active site. If this run looks good it might be worth it to analyze a possible active site at the C-terminus. OPPS! I did not see Kevin's new comments he put about the sheet constraints. I already have the above constraints set up, so I guess its worth running to see the difference in strand alignments, but I'll look at the difference on paper while try6 is running. Try6 running on meow. Mon Aug 9 23:56:12 PDT 2004 Kevin Karplus Aside from needing to pack helix E214-R230 against the barrel, try7-opt2 looks pretty good. Oops, R191 is on the wrong side of the sheet, as is E143. Interestingly, try4-opt2 and try2-opt2 score better than try7-opt2 with the try7 costfcn, which is rather fishy---probably reflecting the hbond constraints violted by the flipped sheets in try7. We may need to decrease the weights for sidechain (which is way overweighted at 6) and increase constraints to get better behavior. It might be worth looking at try7-opt2.constraints, to see which ones are most violated. The worst Hbond is L140.N T111.O---instead I have V142.N-T111.O, so perhaps I have that strand pairing shifted by 2. Both strands appear to be flipped. Maybe E143 should line up with T164? T111 with T70? Y189 with Y235? Tue Aug 10 15:57:23 PDT 2004 Kevin Karplus Bret and I had a discussion about the barrel, and he will try to type up the pattern we'll try next. Tue Aug 10 10:37 2004 Bret Barnes Ok, so I have three strand topologies. The first one is the one Kevin and I discussed today. Which as it turns out is the exact same strand ordering as try7, however I want to see if adding a few constraints to cluster the conserved hydrophobics on the sheet helps correct the structure. Topology 1: oioioioioioioioi 1: MVYVGTS 39: rIVELNF 67: FYFTVK 110: MTLAQF 139: eLAVEFR 164: TFVVVDe 187: YAYFRFH 233: ETYVFFN 1: MVYVGTS The second topology is taken from the sheet constraints that resulted from try7. This is not the best topology on paper, but it might be worth seeing how it works. I also added the missing strand constraints by analyzing the structure. Since this sheet ordering places some of the conserved hydrophilics on the outside of the sheet I'm also going add the same constraints as above to try to cluster them in the center of the barrel. Topology of try7-opt.pdb ioioioioioioioi 3: YVGTs 39: rIVEL 67: fYFTVK 108: lkmTLAQ 139: eLAVEFR 161: hgitFV 187: YAYFRF 223: eTYVFF 3: YVGT Topology 2: (slightly different) ioioioioioioioi 3: YVGTs 39: rIVEL 67: fYFTVK 108: lkmTLAQ 140: LAVEFR 161: hgitF 188: AYFRF 233: eTYVFF 3: YVGT The third topology is based of an improved version of the second topology (from the try7 structure). I want to try this topology first with out the constraints on the hydrophilics to see if it will place them on the inside of the sheet by itself. Topology 3: (Ok, this one I am stil re-thinking, but for right now it is as follows, basically another way to mix up the conserved hydrophilics and still try to keep most of the hydrophobics on the outside) ioioioioioioioi 3: YVGTS 39: RIVEL 67: FYFTVK 110: MTLAQ 142: VEFRH 163: ITFVVV 188: AYFRFH 234: TYVFFN 1: MVYV Wed Aug 11 10:45:09 PDT 2004 Kevin Karplus Bret did all that thinking about sheets, but did not turn ANY of it into cost functions and undertaker runs. There has not been a run started since try7 on Aug 9. This target is due in 25 hours---why are the computers not working on trying to get these sheets??? For try7, I tried to get oioioioioioioioi 1: MVYVGTS 39: rIVELNF 67: FYFTVK 110: MTLAQF 139: eLAVEFR 164: TFVVVDe 187: YAYFRFH 233: ETYVFFN 1: MVYVGTS But what the try7 model created was oioioioioioioioi 1: MVYVGTS 39: rIVELNF 67: FYFTVK 110: MTLAQF 139: eLAVEFR 162: giTFVVVDe 187: YAYFRFH 233: ETYVFFN 1: MVYVGTS Aiming for a compromise that has the conserved polars inside the barrel, I'll try oioioioioioioioi 1: MVYVGTS 39: rIVELNF 67: FYFTVK 110: MTLAQF 139: eLAVEFR 162: giTFVVVDe 187: YAYFRFH 233: ETYVFFN 1: MVYVGTS SheetConstraint G5 S7 I40 E42 hbond T6 SheetConstraint I40 N44 Y68 K72 hbond V41 SheetConstraint T70 K72 M110 L112 hbond V71 SheetConstraint M110 Q114 A141 R145 hbond T111 SheetConstraint V142 R145 T164 V167 hbond V142 SheetConstraint T164 D169 A188 H193 hbond V166 SheetConstraint Y189 H193 E233 F237 hbond F190 SheetConstraint E233 N239 M1 S7 hbond V236 Let's add a salt bridge between E42.CD and R191.CZ and between K72.NZ and E143.CD, to help in orienting the stands. I'll make this try8K.costfcn, since Martina seems to be starting a try8 and Bret a try9. Wed Aug 11 11:21 Bret Barnes I'm removing try9 and letting Martina use that, so that she has try8 and try9. I'll do try10, and try11. Wed Aug 11 11:29:30 PDT 2004 Kevin Karplus Before I start try8K, I'll add a few more TIM-barrel templates to the MANUAL_TOP_HITS and make extra_alignments and all-align* This should give us a few more possibilities for getting a good barrel alignment. Wed Aug 11 11:49:16 PDT 2004 Martina Koeva I have started try8 as proposed above, increasing the weights on the sheet constraints and introducing two additional constraints to try to keep the polar residues R191 and E143 on the inside of the barrel. Have turned down the weight on 'sidechain' by half. Try9 uses the topology that I had proposed earlier. It translates into the following sheet constraints: SheetConstraint V4 S7 R39 E42 hbond T6 10 SheetConstraint R39 L43 F67 V71 hbond E42 10 SheetConstraint T70 K72 M110 L112 hbond V71 10 SheetConstraint L112 A113 L140 A141 hbond L112 10 SheetConstraint A141 F144 T164 V167 hbond A141 10 SheetConstraint T164 V167 A188 R191 hbond F165 10 SheetConstraint A188 R191 E233 V236 hbond F190 10 SheetConstraint T234 V236 V2 V4 hbond V236 10 Once these trys are done, we would need a couple of additional constraints to pack the helix E214-R230 close to the barrel. Wed Aug 11 12:45 2004 Bret Barnes Try10 is running on meow. I used the constraint that I derived from the try7 sheet constraints (Topology 2 above). They go as follows: SheetConstraint V4 S7 R39 E42 hbond V4 7 SheetConstraint R39 L43 F67 V71 hbond I40 7 SheetConstraint Y68 K72 L108 L112 hbond F69 7 SheetConstraint K109 Q114 L140 R145 hbond M110 7 SheetConstraint V142 R145 H161 F165 hbond V142 7 SheetConstraint H161 F165 A188 F192 hbond G162 7 SheetConstraint Y189 F192 E233 V236 hbond F190 7 SheetConstraint Y235 F238 Y3 T6 hbond Y135 7 I also added bonus constraints to cluster the conserved hydrophobics in the center of the barrel. I determined the distances by analyzing try7 distances. Constraint E42.CD K72.NZ 3.0 4.535 6.5 1.0 bonus Constraint K72.NZ R145.CZ 4.0 8.0 12.0 1.0 bonus Constraint K72.NZ E143.CD 3.0 5.647 7.5 1.0 bonus Constraint E143.CD R145.CZ 4.0 8.0 12.0 1.0 bonus Constraint K72.NZ Q114.CD 3.0 5.534 6.5 1.0 bonus Constraint Q114.CD R145.CZ 5.0 7.0 10.0 1.0 bonus Constraint E42.CD R191.CZ 3.0 5.647 7.5 1.0 bonus Constraint Q114.CD R191.CZ 4.5 8.0 10.0 1.0 bonus For try11 I used the sheet constraints from topology 3 (above). I came up with these by modifying the sheet constraints produced by try7. I also included the bonus constraints I put in try10 to try to cluster the conserved hydrophobics. Sheet Constraints for try11: SheetConstraint V4 S7 R39 E42 hbond V4 7 SheetConstraint R39 L43 F67 V71 hbond I40 7 SheetConstraint Y68 K72 M110 Q114 hbond F69 7 SheetConstraint T111 Q114 V142 R145 hbond L112 7 SheetConstraint V142 H146 I163 V167 hbond F144 7 SheetConstraint I163 V168 A188 H193 hbond V166 7 SheetConstraint Y189 H193 T234 F238 hbond F192 7 SheetConstraint V236 N239 1M V4 hbond F238 7 try11 is running on tweet. Wed Aug 11 13:46 2004 Bret Barnes Looking at try7 it seems that the last helix (214-230) should pack against the open hydrophobic gap (strand 7, residues 188-192). The helix's hydrophobic residues are 228,227,224,223,221,220,217. F221 on the helix could pack well inbetween the two phenylalanines on the strand F190 and F192. That is of course if this strand is actually the correct strand. Wed Aug 11 14:32 2004 Bret Barnes try8K-opt1 still does not flip all the strands in the correct orientation. However strand 7 is still exposed (even though it does need to be flipped) and it positions F190 and F192 in the correct spot to interact with F221 of the last helix (except they are facing into the barrel). Wed Aug 11 14:51:57 PDT 2004 Kevin Karplus Foo, even with all the extra constraints, try8K-opt2 still has R191 and E143 facing out of the barrel instead of into it. Also, try7-opt2 still scores better with the try8K costfcn. The try9 costfcn also prefers try7-opt2 (try9-opt1 and try10-opt1 are way down on the list). The try10 costfcn prefers try8K-opt2, but try10-opt2 is not too far down the list, so improvement is possible. All of these have the same problems with R191 and E143 facing out of the barrel. Perhaps I should take the best structural alignment for try7-opt2 and edit it by hand to move those strands over 1. VAST job: VS60675 Password: T0217try7 So once I have the VAST alignment, how should I hand edit it? Move the segments 138: yeLAVEFRh back 1 so it lines up one residue earlier 162: giTFVVVDe back 3 187: YAYFRFH back 1 Then optimize using just that hand alignment as a starting point for the TryAllAlign (the other alignments can be used for the fragment library still). Wed Aug 11 15:41:22 PDT 2004 Martina Koeva Given everything that has been done so far in trying to flip the strands, I think it would be worth a try to just keep R191 and E143 flipped out and try to salt bridge them to residues on a helix. I am setting up a costfcn (as a try12) that will use the sheet constraints from try7 and try to pack the helix 213-230 down closer to the barrel. I am introducing an additional constraint that pairs up residues R191 and E216. Constraint R191.CZ E216.CD -10 4.0 6.0 7 Constraint E143.CD R151.CZ -10 4.0 6.0 7 Meanwhile, try8 has crashed and I had to restart it again. Wed Aug 11 16:36 2004 Bret Barnes E143 is right next to R151 and H146 which might make it ok to leave it on the outside of the barrel. R191 on the other hand really should be stuffed into the barrel. The strand alignment produced by try7 could be fixed if you just moved strand 7 one foward or one backward. That would flip the strand and put 191 on the right side. I'm going to try to use the exact sheet constraints that try7 produced, but move strand 7 one forward (try13) and one backward (try14). Also I'm going to add a constraint to put R191 close to E42. Wed Aug 11 16:58:21 PDT 2004 Kevin Karplus The VAST run has finally finished, and it like the alignment to 1pz1A best. I'll save the alignment as 1pz1A/1pz1A-T0217-try7-vast.a2m. The edited alignment will be 1pz1A/1pz1A-T0217-hand1.a2m. I'll start run try15 optimizing with just that hand alignment as a starting point, and InsertAlignment turned off, so no other alignment can creep in. Wed Aug 11 18:03 2004 Bret Barnes try13 running on baa. try14 running on cluck.From karplus@soe.ucsc.edu Wed Aug 11 18:07:18 2004 Date: Wed, 11 Aug 2004 18:07:16 -0700 From: Kevin Karplus To: martina@soe.ucsc.edu, bbarnes@soe.ucsc.edu CC: karplus@soe.ucsc.edu Subject: submit tonight I'm going to put together a submission tonight for T0217. If there is anything better in the morning we can resubmit. Right now, it looks like try7-opt2 is our best hit. We can submit 5, and I currently plan try7-opt2 best scoring try9-opt2.repack-nonPC best Rosetta energy try1-opt2 full auto T0217-1oltA-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 (first in undertaker-align) That leaves room for one more model. Which one would YOU like to see included? ------------------------------------------------------------ Wed Aug 11 19:51:17 PDT 2004 Kevin Karplus Not hearing any comments, I'll look at the score-all.try15.rdb file, which orders them try9-opt2, try15-opt1, try10-opt2, try7-opt2. I'd better look at the new files, in case one of them is actually any good. The unconstrained costfcn prefers try3-opt2, try8K-opt2, try6-opt2, try9-opt2.repack-nonPC, ... YES! try15-opt1 has the strands oriented the right way! The problem must have been that there were no good alignments in the all-align set (or that there were bad alignments that scored better). Neither try9 nor try10 fix the bad strands. I'm going to put try15-opt1 first (and replace it later by try15-opt2). I just noticed that "unconstrained.costfcn" still had a high weight for sidechain. I'll twiddle the weights a little bit to be more reasonable. Now the order is try9-opt2.repack-nonPC, try3-opt2, ... Wed Aug 11 20:13:50 PDT 2004 Kevin Karplus submit: try15-opt2 desired orientation of strands try15-opt2.repack-nonPC repacked by Rosetta try9-opt2.repack-nonPC best Rosetta energy, lowest unconstrained cost try1-opt2 full auto T0217-1oltA-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 (first in undertaker-align) I might try to reoptimize try15-opt2, now that we have it. ============================== From martina@soe.ucsc.edu Wed Aug 11 20:02:36 2004 MIME-Version: 1.0 Date: Wed, 11 Aug 2004 20:02:34 -0700 (PDT) From: Martina Koeva To: Kevin Karplus cc: martina@soe.ucsc.edu, Subject: Re: submit tonight In-Reply-To: <200408120107.i7C17GHc006560@cheep.cse.ucsc.edu> Yay! Try15-opt1 has both E143 and R191 on the inside of the barrel . It has unwound the helix 213-230 a bit, but otherwise the helix is packed quite a bit closer to the barrel than in try7. Martina From martina@soe.ucsc.edu Wed Aug 11 20:17:56 2004 MIME-Version: 1.0 Date: Wed, 11 Aug 2004 20:17:54 -0700 (PDT) From: Martina Koeva To: Kevin Karplus cc: martina@soe.ucsc.edu, Subject: Re: submit tonight In-Reply-To: <200408120107.i7C17GHc006560@cheep.cse.ucsc.edu> I was looking at score-all.try15.rdb file, which seems to like try6-opt2.repack-nonPC better than the one from try9. It also seems to like (if I am not wrong) try9-opt2 slightly better than try8K-opt2. So I am proposing currently: try15-opt2 try6-opt2.repack-nonPC try9-opt2 try1-opt2 T0217-1oltA-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 Martina ============================================================ From karplus@soe.ucsc.edu Wed Aug 11 20:29:29 2004 Date: Wed, 11 Aug 2004 20:29:28 -0700 From: Kevin Karplus To: martina@soe.ucsc.edu CC: karplus@soe.ucsc.edu, bbarnes@soe.ucsc.edu In-reply-to: (message from Martina Koeva on Wed, 11 Aug 2004 20:17:54 -0700 (PDT)) Subject: Re: submit tonight For the repack-nonPC model, I like to submit the one with the lowest energy reported by Rosetta, which is try9-opt2.repack-nonPC. See the README file for the set of predictions I currently plan to submit. I could replace try15-opt2.repack-nonPC with try6-opt2.repack-nonPC if it gives us a better set of predictions. ============================================================ Wed Aug 11 20:58:31 PDT 2004 Kevin Karplus I'm starting a try16 to optimize try15-opt2 a bit more. If it works, which it should, I'll replace the try15-opt2 models in the morning. Thu Aug 12 00:57:15 PDT 2004 Martina Koeva Should the new set of models be: try16-opt2 try16-opt2.repack-nonPC try9-opt2.repack-nonPC try1-opt2 T0217-1oltA-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5? Try16-opt2 does score better than try15-opt2, as does try16-opt2.repack-nonPC with the try16 cost function. There is still a break in the backbone around residues 233-234. Would it be good idea to try and run an additional try in an attempt to close the gap? Wed 12 1:05 2004 Bret Barnes So from what I notice the helix did not pack against the barrel very well. I'm going to try running try15 over again, but pack F200 against F192 and see what happens. try17 running on meow. Thu Aug 12 01:30:25 PDT 2004 Martina Koeva I have increased the break weight and some of the dry weights for try18 and have started it on hoot.I have otherwise kept the .costfcn the same as the one from try16. Thu Aug 12 04:35:03 PDT 2004 Kevin Karplus I resubmitted try16-opt2 try16-opt2.repack-nonPC try9-opt2.repack-nonPC try1-opt2 T0217-1oltA-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 fixing up the methods for the first three models. try9-opt2.repack-nonPC is still the best Rosetta energy function, but no longer the best with unconstrained.costfcn (try16-opt2 is now). Since try16-opt2 now scores best with unconstrained.costfcn, I'll reoptimize using that (try19). SOMEONE messed around with try16.under AFTER it was run. THIS IS A NO-NO. NEVER CHANGE A .under file after it has run, as this destroys the record of what was actually done. Even if no changes are made, simply changing the date and time of the file destroys confidence in the record. It also means that an accidental rerun of "make" could rerun the script and destroy the output, as the output is no longer newer than the script. Thu Aug 12 09:29:46 PDT 2004 Kevin Karplus try19-opt2 is the new best scoring, but I'm a bit worried about T111 and E139. Both seem to be on the wrong side of the barrel. The T111 error is probably due to a flipped strand that I did not catch earlier, since both T111 and K109 should be inside the barrel. (R107 and K109 are getting a reasonable orientation by twisting and bending the strand.) The incorrect orientation of the strand containing T111 has been there since try15.costfcn---ah, we did that to make the conserved Q114 be on the right side. Hmm, I don't see a simple solution---we could slide the strand forward 3, redefining where we think the strand is, but that would require another full day of optimization. My temptation is to submit try19-opt2 and let it go. Thu May 26 15:22:41 PDT 2005 Kevin Karplus I picked up the correct structure for T0217 today and am evaluating it for the CASP6 paper. The differences between the models we were considering for submission is rather tiny---most of our models score 25-27 in smooth GDT.