Mon Jul 12 10:45:48 PDT 2004 T0234 DUE 11 Aug Mon Jul 12 18:21:03 PDT 2004 Kevin Karplus Comparative model b.45.1.1 (1dnlA). The initial alignments look good, but the helix and strand constraints seem to mess stuff up in try1. For try2 we should pick up the actual helices and strands found in the best alignments, and reoptimize. (Oops, forgot to comment out PrintTemplates in try2---be sure to do it in try3.) Sat Jul 17 18:12:27 PDT 2004 Kevin Karplus try2 has much of the model ok, but the N-terminal helix needs to pack against the sheet. Perhaps a constraint for F14.CZ and L87.CG would help pin the N-terminal helix in place. The last strand G144-S150 may need straightening. Actually, I'm not so sure it is a strand---the helix nicely points the hydrophobics inward. Maybe I'll stick with the alignments on this one. Sat Jul 17 22:04:32 PDT 2004 Kevin Karplus try3-op2 is looking fairly good. There is a big hydrophobic patch that looks like a multimerization interface. (Possibly a 4mer or 5mer?) The main hit (1dnlA) is part of a dimer. The N terminal helix may provide a looser connection that we assumed for try3 (see 1dnlA). The residues K102 to F127 seem out of place---where should the helix here go? Sun Jul 18 08:49:12 PDT 2004 Kevin Karplus I made the dimer from the "biological unit" from PDB (see the fetch_dimer_1dnlA target of the Makefile and make-dimer-try3.under). It looks pretty good except for the loose helix, which looks like it should rotate to pack F110 and F114 against V40, and I117 against A37 or Y36. I suspect that sheet constraint for P38 A42 V51 A47 hbond I41 would also help pull the helix down against the barrel. This is just a lengthening of the existing sheet constraint from try3-opt2.sheets I've set up a try4 cost function and will try optimizing again from the alignments. Sun Jul 18 11:11:08 PDT 2004 Kevin Karplus The try4 optimization did not score as well with the try4 cost function as try3 did. Looking at the models from alignments, I'm not so sure I WANT to pack the helix with F110 and F114 where I've been trying to. Maybe I should just relax all the constraints and repack everything from the existing models? Turning the dry weights way up may help a bit, though without constraints it will be hard for the "Opt" operators to make big moves. Sun Jul 18 15:01:52 PDT 2004 Kevin Karplus Other than the problems with the helix containing F110 and F114, try5-opt2 looks pretty good. There are breaks at each end of the helix, so if I drop the break cost way down, then OptSegment SHOULD be able to place it where I want, if I use strong constraints. Let's try once more to fill the huge hole. I'll add the constraints, reduce soft_clashes and breaks, then adjust the weights of other components of the cost function until try5-opt2 is just barely the best. If we do get the helix to turn, we'll need to reoptimize with breaks and clashes turned back up. I'm afraid that try2 will take over, though since it scores almost as well as try5 on these measures, and is much easier to fix. I'll simulataneously do a try7 with the same cost function, but starting ONLY from try3-opt2, try4-opt2, and try5-opt2. Sun Jul 18 21:01:56 PDT 2004 Kevin Karplus try6-opt2 is the new best score, but try7-opt2 is not far behind. Rosetta hates the repacked versions of try6 and try7 though, probably because of chain breaks or clashes. In try6, the helix we were having trouble with has packed in fairly well, but the N-terminal helix is sticking way out. In try7, the helix has turned, but is not packed in tightly, and the gaps are ridiculoously large. I'll turn up the clashes and breaks and reoptimize from existing models. Before doing the optimization, I made a chimeric model from try6 and try7, copying I96-F127 from try6 into try7. This chimera scores the best with the try8 cost function, which has constraints for both the helices. After polishing it up, I'll try reoptimizing without constraints (or with weak constraints, if necessary). Mon Jul 19 07:57:03 PDT 2004 Kevin Karplus try8 looks fairly good, but there needs to be tighter packing for F114 and F110. For try9, I'll change the constraints to try to pull those in tighter, and remove other constraints. In the dimer, there are some clashes between T161-G165 at the Cterminal end and R113-V120. Slightly tighter packing of the helix would alleviate this clash, without having to pin back the C-terminus, but extending the final sheet constraint out to F143 would probably help: SheetConstraint F143 G151 G162 L154 hbond Y147 Mon Jul 19 10:42:31 PDT 2004 Kevin Karplus try9 is looking better. The broken helix could be better packed against the sheet. There are a few conflicts in the dimer, but they could probably be fixed by repacking. Mon Jul 19 21:20:49 PDT 2004 Kevin Karplus After making some changes to the make-repack-res-file scripte, I made a repacked version of the dimer-1ndlA-try9 file. The initial file had 77 atoms of chain A within 3.5Ang of chain B, and 20 within 2.0 Ang. The repacked file had 200 atoms within 3.5 but only 15 within 20. I submitted ReadConformPDB T0234.try9-opt2.pdb ReadConformPDB T0234.try8-opt2.repack-nonPC.pdb # best Rosetta energy ReadConformPDB T0234.try9-monomer-from-repacked-dimer.pdb ReadConformPDB T0234.try1-opt2.pdb ReadConformPDB T0234.t2k.undertaker-align.pdb model 1 Sun Sep 19 19:23:17 PDT 2004 Kevin Karplus I added REAL_PDB:=1vl7A FINAL_COSTFCN:=try9 to the Makefile and evaluated the models using whole-chain rmsd. We did pretty good on this one. The order of submitted and robetta models was model1, model3, model2, model5, robetta4, robetta5, robetta1, robetta3, robetta2, model4. The fully automatic model4 is awful (at 16.34) but the best model is pretty good at 5.526 Ang. We did submit our very best model as model1. Rosetta repacking made try8-opt2 worse, and even repacking as a dimer didn't help. Fri Sep 24 21:20:25 PDT 2004 Kevin Karplus Using the new smooth_GDT, we get name length missing_atoms rmsd rmsd_ca GDT smooth_GDT T0234.try6-opt1.pdb.gz 165 0.0000 6.0327 4.9851 -57.9630 -54.8164 best with smooth_GDT model3.ts-submitted 165 0.0000 5.5946 4.3641 -62.0370 -54.6949 from dimer model1.ts-submitted 165 0.0000 5.5261 4.3642 -59.0741 -54.5706 try9 model2.ts-submitted 165 0.0000 5.6187 4.3602 -59.0741 -54.5247 try8-repacked robetta-model1.pdb.gz 165 0.0000 7.4576 6.5650 -57.7778 -54.2386 robetta-model4.pdb.gz 165 0.0000 7.0737 6.1654 -57.0370 -54.2384 robetta-model3.pdb.gz 165 0.0000 7.7905 6.8235 -56.2963 -53.6314 robetta-model5.pdb.gz 165 0.0000 7.2753 6.4563 -56.1111 -53.6013 model5.ts-submitted 165 240 6.2721 4.9441 -56.2963 -52.2096 alignment robetta-model2.pdb.gz 165 0.0000 7.9970 7.1800 -54.2593 -51.4890 model4.ts-submitted 165 0.0000 16.3424 16.0432 -25.7407 -24.9186 full auto