Thu Jul 22 10:50:01 PDT 2004 T0247 DUE 13 Aug Thu Jul 22 14:16:25 PDT 2004 Kevin Karplus This looks like a comparative model with 1pj5A as the template. Thu Jul 22 20:41:09 PDT 2004 Kevin Karplus The 1pj5A template is longer than the target, and seems to cover all of it. The task here will mainly be one of picking an alignment, and building very short loops. The try1-opt2 alignment looks pretty good---some polishing to close gaps and tighten packing may be all it needs. The rr constraints look fairly good, but they're not really needed on this target. Fri Jul 23 11:36:16 PDT 2004 Kevin Karplus try2 looks pretty good and has reduced clashes and breaks. I might want to add a couple of sheet constraints: E309-T313 isn't forming the hbonds of a barrel, but probably should. sheetconstraint G310 I312 V330 A328 hbond R329 sheetconstraint E309 G310 R299 V298 hbond V298 The hairpin at A16-L28 probably should attach to the first sheet. It isn't coming from the template. The edge strand of the current sheet is G45-D48, with M46 free to Hbond. Somewhere in W24-Y30 is the matching strand, unless it is a loop, and R17-F21 is the matching strand. Let's try sheetconstraint R17 D20 G45 D48 hbond M46 Fri Jul 30 00:10:17 PDT 2004 Kevin Karplus try3 looks a bit foamy, as if the clash term was turned up too high relative to the other terms. There are still some bad breaks: T0247.try3-opt2 breaks before (T0247)H29 with cost 6.77975 T0247.try3-opt2 breaks before (T0247)K280 with cost 4.5625 T0247.try3-opt2 breaks before (T0247)G335 with cost 3.50854 T0247.try3-opt2 breaks before (T0247)P218 with cost 3.48768 T0247.try3-opt2 breaks before (T0247)K79 with cost 3.38699 T0247.try3-opt2 breaks before (T0247)D140 with cost 3.19492 T0247.try3-opt2 breaks before (T0247)G290 with cost 2.75513 T0247.try3-opt2 breaks before (T0247)K289 with cost 2.40644 T0247.try3-opt2 breaks before (T0247)P131 with cost 2.31966 T0247.try3-opt2 breaks before (T0247)R262 with cost 2.26677 T0247.try3-opt2 breaks before (T0247)Q307 with cost 2.22222 T0247.try3-opt2 breaks before (T0247)F264 with cost 1.49627 T0247.try3-opt2 breaks before (T0247)V168 with cost 1.49071 T0247.try3-opt2 breaks before (T0247)H276 with cost 1.21895 T0247.try3-opt2 breaks before (T0247)P173 with cost 1.04073 T0247.try3-opt2 breaks before (T0247)A91 with cost 1.01574 For try4, try increasing break weight and dry weights, but reduce soft_clashes. Maybe increase hbond weights?? Fri Jul 30 13:52:07 PDT 2004 Sol Katzman Try4 looks just as foamy to my eyes, and has about the same set of breaks. T0247.try4-opt2.pdb.gz breaks before (T0247)H29 with cost 6.38033 T0247.try4-opt2.pdb.gz breaks before (T0247)G335 with cost 3.47563 T0247.try4-opt2.pdb.gz breaks before (T0247)K79 with cost 3.38432 T0247.try4-opt2.pdb.gz breaks before (T0247)D140 with cost 3.08965 T0247.try4-opt2.pdb.gz breaks before (T0247)P218 with cost 3.04513 T0247.try4-opt2.pdb.gz breaks before (T0247)G290 with cost 2.49292 T0247.try4-opt2.pdb.gz breaks before (T0247)K289 with cost 2.31317 T0247.try4-opt2.pdb.gz breaks before (T0247)R262 with cost 2.29837 T0247.try4-opt2.pdb.gz breaks before (T0247)P131 with cost 2.13324 T0247.try4-opt2.pdb.gz breaks before (T0247)Q307 with cost 1.87059 T0247.try4-opt2.pdb.gz breaks before (T0247)F264 with cost 1.51392 T0247.try4-opt2.pdb.gz breaks before (T0247)V168 with cost 1.49996 T0247.try4-opt2.pdb.gz breaks before (T0247)A91 with cost 1.07968 T0247.try4-opt2.pdb.gz breaks before (T0247)P173 with cost 1.04832 Clashes were REDUCED from try3->try4 (274->240) but Rosetta actually likes try2 (which only had 164 clashes) much better than either try3 or try4 after repacking. I believe that all-align.a2m.gz was missing prior to try4 so that may account from some of the difference. (Also note that the SheetConstraints from try3-opt2 were used in try4.) At this point, I am content to wait and see the master's next move, so I defer to Kevin to continue. Sat Jul 31 04:35:23 PDT 2004 Kevin Karplus If all-align was missing, then maybe I should do another run from the alignments, using something like the try2 costfcn. That may pick up some better stuff than trying to fix up what we have so far. I've got too many targets to look at, so after try5, I think I'll leave this in Sol's hands. Mon Aug 2 11:18:24 PDT 2004 Kevin Karplus The try5 costfcn likes try2, try4, try3, and try1 all more than it likes try5. The region P246-H276 looks bad, but the rest seems ok, aside from the problem that everything is a bit foamy. Maybe the thing to do is to turn up the dry weights (particularly dry5 and dry6.5) and just polish the existing models. I'll leave this to Sol to do. Mon Aug 2 16:03:26 PDT 2004 Sol Katzman For try6, put back the SheetConstraints from try3, read in all previous tries, skip the TryAllAlign stuff, and up the dry weights a little (they were already fairly high). Tue Aug 3 15:22:53 PDT 2004 Sol Katzman The try6 costfcn likes try6 best, then the rest in the order of try5 costfcn. For the template 1nrkA (which at 328 residues is easier to align to the 364 residues of T0247 than 1pj5A with its 830 residues) there is an SO4 ion binding site with these residues within 3Ang: Atom: ND2 1524 Group: ASN 193 Chain: A Atom: NE1 216 Group: TRP 27 Chain: A Atom: NE1 1490 Group: TRP 189 Chain: A Atom: CD1 1282 Group: LEU 162 Chain: A We can use these distances as constraints on the binding site, as measured in rasmol on 1nrk: Distance TRP27A.NE1-ASN193A.ND2: 4.881 Distance TRP27A.NE1-TRP189A.NE1: 5.412 Distance TRP27A.NE1-LEU162A.CD1: 6.758 Distance ASN193A.ND2-TRP189A.NE1: 3.663 Distance ASN193A.ND2-LEU162A.CD1: 7.406 Distance TRP189A.NE1-LEU162A.CD1: 4.730 I make the following associations from 1nrkA to T0247 roughly based on alignment T0247-1nrkA-t04-global-str2+CB_burial_14_7-0.4+0.4-adpstyle5: 1nrk T0247 ---- ----- W27 H51 L162 E196 W189 R224 N193 R228 Leading to these constraints on T0247: Constraint H51.NE2 R228.NH2 1.881 4.881 7.881 Constraint H51.NE2 R224.NH2 2.412 5.412 8.412 Constraint H51.NE2 E196.OE2 3.758 6.758 9.758 Constraint R228.NH2 R224.NH2 0.663 3.663 6.663 Constraint R228.NH2 E196.OE2 4.406 7.406 10.406 Constraint R224.NH2 E196.OE2 1.730 4.730 7.730 For try7, include the above as well as weight 20 on constraints NOTE: the weight for constraints was previously absent. Also the script pointed to by 'bind' defines the residues as "bind" and "bind1nrk" (and also "so41nrk"). The script pointed to by 'dobind' will execute the 'bind' script, then spacefill and color-cpk the bind residues. Thu Aug 5 13:01:14 PDT 2004 Sol Katzman It looks like try7 did bring the active atoms of the supposed binding residues closer together. It definitely increased the number of clashes compared to try6 (156 -> 192), although this is still better than the other tries (except try2 at 164 clashes). The try7 costfcn likes try7,try6, then try2,try1,try3. Rosetta likes repack-nonPC try2 the best, then try5, try6, try7. For try8, increase weight for soft_clashes (20 -> 30) and decrease weight for constraints (20 -> 10). Also, include rr.constraints, (those above 0.65 weight) with option 'bonus'. Mon Aug 9 10:29:46 PDT 2004 Sol Katzman Rosetta likes try8 second best now (after try2), then try5,try6,try7 The try8 costfcn likes try8,try7,try6,try2. The breaks are not too bad, nor are the clashes. Conformation[30] T0247.try8-opt2.pdb.gz has 78 breaks (sorted:) T0247.try8-opt2.pdb.gz breaks before (T0247)G335 with cost 3.47076 T0247.try8-opt2.pdb.gz breaks before (T0247)K79 with cost 3.37377 T0247.try8-opt2.pdb.gz breaks before (T0247)P131 with cost 2.72053 T0247.try8-opt2.pdb.gz breaks before (T0247)G290 with cost 2.41698 T0247.try8-opt2.pdb.gz breaks before (T0247)K289 with cost 2.29249 T0247.try8-opt2.pdb.gz breaks before (T0247)Q307 with cost 1.61718 For conformation number 30 T0247.try8-opt2.pdb.gz: T0247 has 148 clashes, 0 disulphide bonds, and 248 hydrogen bonds in 364 residues other-bump: 1.62858 Ang (T0247)L292.CD2 and (T0247)I342.CD1 threshold= 3.03525 cost= 0.976139 other-bump: 1.58347 Ang (T0247)Y85.O and (T0247)A364.CB threshold= 2.77262 cost= 0.965302 other-bump: 1.7888 Ang (T0247)L57.CD2 and (T0247)I136.CD1 threshold= 3.03525 cost= 0.958101 other-bump: 2.09651 Ang (T0247)M88.SD and (T0247)R118.NH2 threshold= 3.27662 cost= 0.931384 other-bump: 1.89399 Ang (T0247)T313.OG1 and (T0247)L327.CB threshold= 2.82678 cost= 0.909529 other-bump: 2.17969 Ang (T0247)T53.CG2 and (T0247)V113.CG2 threshold= 3.06375 cost= 0.870328 other-bump: 2.23009 Ang (T0247)I34.CG1 and (T0247)P319.CB threshold= 3.07242 cost= 0.853764 Mon Aug 9 17:14:09 PDT 2004 Kevin Karplus I made an unconstrained.costfcn, which scores try8-opt2 best. The top 5 or 6 breaks in try8-opt2 are still pretty bad. I'm going to do a polishing run without constraints to see if we can close them up a little. Tue Aug 10 00:25:27 PDT 2004 Kevin Karplus try9-opt still has bad breaks, but not quite as bad: Conformation[46] T0247.try9-opt2 has 75 breaks T0247.try9-opt2 breaks before (T0247)K79 with cost 3.37452 T0247.try9-opt2 breaks before (T0247)P131 with cost 2.65998 T0247.try9-opt2 breaks before (T0247)G290 with cost 2.4261 T0247.try9-opt2 breaks before (T0247)K289 with cost 2.29988 T0247.try9-opt2 breaks before (T0247)V168 with cost 1.2881 T0247.try9-opt2 breaks before (T0247)P218 with cost 1.26628 T0247.try9-opt2 breaks before (T0247)F264 with cost 1.09626 There are also still bad clashes, though not quite as bad as before. For conformation number 39 T0247.try9-opt2.pdb.gz: T0247 has 130 clashes, 0 disulphide bonds, and 253 hydrogen bonds in 364 residues other-bump: 2.12085 Ang (T0247)M88.SD and (T0247)R118.NH2 threshold= 3.27662 cost= 0.926463 other-bump: 1.84582 Ang (T0247)Y85.O and (T0247)A364.CB threshold= 2.77262 cost= 0.912944 other-bump: 1.90871 Ang (T0247)T313.OG1 and (T0247)L327.CB threshold= 2.82678 cost= 0.905225 other-bump: 2.20115 Ang (T0247)T53.CG2 and (T0247)V113.CG2 threshold= 3.06375 cost= 0.862477 other-bump: 2.33613 Ang (T0247)L182.CD1 and (T0247)F207.CE1 threshold= 3.11884 cost= 0.823386 other-bump: 2.38927 Ang (T0247)I34.CG1 and (T0247)P319.CB threshold= 3.07242 cost= 0.778841 other-bump: 2.51964 Ang (T0247)K82.N and (T0247)Y103.CE2 threshold= 3.1566 cost= 0.74135 Rosetta still likes try2-opt2.repack-nonPC best, but try9-opt2 is second best. I don't think there is much point in doing further polishing. We could probably reduce the breaks and clashes somewhat, if we ran long enough, but the differences would be small. I'll submit try9-opt2 undertaker likes best try9-opt2.repack-nonPC try2-opt2.repack-nonPC Rosetta likes best try1-opt2 full auto T0247-1pj5A-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 Fri Sep 24 21:44:53 PDT 2004 Kevin Karplus Evaluating using smooth_GDT, we get name length missing_atoms rmsd rmsd_ca GDT smooth_GDT model5.ts-submitted 364 92. 4.4635 3.6618 -64.3352 -59.8487 alignment robetta-model3.pdb.gz 364 0.0000 4.6866 3.9656 -61.7036 -58.3175 robetta-model1.pdb.gz 364 0.0000 4.8615 4.2372 -62.5346 -57.6830 robetta-model2.pdb.gz 364 0.0000 5.5213 4.9298 -59.9030 -56.2873 model4.ts-submitted 364 0.0000 4.9630 4.4702 -60.8726 -55.9637 full auto robetta-model5.pdb.gz 364 0.0000 5.2132 4.5054 -60.3878 -55.4669 robetta-model4.pdb.gz 364 0.0000 5.3327 4.5438 -58.8643 -54.9911 model3.ts-submitted 364 0.0000 5.0653 4.5524 -58.0332 -54.2485 try2-repack model1.ts-submitted 364 0.0000 6.1049 5.5088 -54.6399 -50.8060 try9 model2.ts-submitted 364 0.0000 6.1451 5.5089 -54.5706 -50.8061 try9-repack Foo, we did best with the alignment. All the predictions are rather similar in quality, but robetta has a slight edge over us.