Fri Jun 2 09:25:27 PDT 2006 T0313 Make started Fri Jun 2 09:26:41 PDT 2006 Running on camano.cse.ucsc.edu Fri Jun 2 11:33:30 PDT 2006 Kevin Karplus comparative model with strong full-length BLAST hits (1sdmA 43%id on 321 residues, Evalue 3.4e-69) 26 BLAST hits with E-value < 1e-10 hits are to c.37.1.9, with 26 strong hits in in w0.5 searches also. The t06 library only has 5 of the templates: 1bg2, 1x88A, 1vfvA, 1f9vA, 1ry6A. Of these, blast sees 1f9vA as the closest, though not quite as good as 1sdmA and (3kar, 1f9tA, 2ncdA, 1cz7A) Fri Jun 2 18:09:25 PDT 2006 Kevin Karplus Aside from V237-H252, try1-opt1 looks pretty good. When everything is done for try1-opt2, I'll have to look at how it compares with the template models. Fri Jun 16 15:55:03 PDT 2006 Kevin Karplus V237-H252 is an insertion relative to all the template alignments. Otherwise the agreement with the templates is excellent. With the unconstrained costfcn, we score best, then the SAM_T06_server_TS1 model. Of the models that are not ours, the best scoring are ROBETTA_TS5 Pmodeller6_TS2 ROBETTA_TS3 HHpred2_TS1-scwrl HHpred3_TS1-scwrl Pmodeller6_TS1 ROBETTA_TS4 3dpro_TS1-scwrl I should look at these also. Sun Jun 18 14:49:51 PDT 2006 Kevin Karplus The agreement with the server models is generally pretty good. There are some disagreements about the hairpin at N23-H36. The o_notor2 predictions put the highest probability of a sep3 N-O bond at D28.O, next highest at F27.O The n_notor2 predictions put the highest probability of a sep3 N-O bond at D31.N, which is consistent. I created try2.costfcn with just one constraint (the Hbond) and am using it to create a model from alignments. (try2 running on whidbey) Mon Jun 19 15:11:06 PDT 2006 Kevin Karplus try2 did not manage to form the H-bond. I'll probably have to add a distance constraint. Tue Jun 20 10:45:17 PDT 2006 Kevin Karplus try3 had the distance for D28.O, D31.N just right at 3 Ang, but did not get a great Hbond score. D28.O is forming a double-Hbond with D32.N and S33.N---there is no room for an Hbond with D31.N. So it looks like trying to force in the Hbond is *not* going to work for me, at least, not without a bunch of other sheet constraints. Tue Jun 20 11:10:43 PDT 2006 Kevin Karplus Despite the failure to do what we wanted, try3-opt2 does score best with the unconstrained costfcn. It ends up with the hairpin in a different place from all the other models, and with the helix V175-R189 at a slightly different angle. I'll try one more run from alignments, with Blast's favorites (1sdmA, 3kar, 1f9tA) as the only source of full-length alignments. I'll do a make extra_alignments and make read_alignments to ensure that the read scripts are all set up correctly. I'll use *no* constraints on try4, to see what undertaker chooses without constraints. (try4 started on orcas) Tue Jun 20 15:23:43 PDT 2006 Kevin Karplus try4 got the same sheets as try3. I'm now not sure what I was trying to do around D28. There seem to be several different opinions about what happens there: Pmodeller6_TS1.sheets: SheetConstraint (T0313)V25 (T0313)D28 (T0313)I35 (T0313)D32 hbond (T0313)T26 1 Pmodeller6_TS2.sheets: SheetConstraint (T0313)V25 (T0313)D28 (T0313)I35 (T0313)D32 hbond (T0313)T26 1 ROBETTA_TS3.sheets: SheetConstraint (T0313)V25 (T0313)D28 (T0313)I35 (T0313)D32 hbond (T0313)T26 1 ROBETTA_TS4.sheets: SheetConstraint (T0313)V25 (T0313)D28 (T0313)I35 (T0313)D32 hbond (T0313)T26 1 ROBETTA_TS5.sheets: SheetConstraint (T0313)A24 (T0313)A29 (T0313)L37 (T0313)D32 hbond (T0313)T26 1 try3-opt2.sheets: SheetConstraint (T0313)F27 (T0313)A29 (T0313)I34 (T0313)D32 hbond (T0313)D28 1 try3-opt2.sheets: SheetConstraint (T0313)E20 (T0313)V25 (T0313)K40 (T0313)I35 hbond (T0313)A21 1 try4-opt2.sheets: SheetConstraint (T0313)F27 (T0313)A29 (T0313)I34 (T0313)D32 hbond (T0313)D28 1 try1-opt2.sheets: SheetConstraint (T0313)A24 (T0313)F27 (T0313)I34 (T0313)D31 hbond (T0313)V25 1 try2-opt2.sheets: SheetConstraint (T0313)V25 (T0313)F27 (T0313)S33 (T0313)D31 hbond (T0313)V25 1 There are also differences of opinion on another strand: Pmodeller6_TS2.sheets: SheetConstraint (T0313)I152 (T0313)R153 (T0313)V164 (T0313)Y163 hbond (T0313)R153 1 ROBETTA_TS3.sheets: SheetConstraint (T0313)I152 (T0313)R153 (T0313)V164 (T0313)Y163 hbond (T0313)R153 1 ROBETTA_TS5.sheets: SheetConstraint (T0313)I152 (T0313)C155 (T0313)V164 (T0313)Q161 hbond (T0313)R153 1 Pmodeller6_TS1.sheets: SheetConstraint (T0313)L154 (T0313)D157 (T0313)V164 (T0313)Q161 hbond (T0313)C155 1 ROBETTA_TS4.sheets: SheetConstraint (T0313)L154 (T0313)D157 (T0313)V164 (T0313)Q161 hbond (T0313)C155 1 try1-opt2.sheets: SheetConstraint (T0313)L154 (T0313)D157 (T0313)V164 (T0313)Q161 hbond (T0313)C155 1 try3-opt2.sheets: SheetConstraint (T0313)L154 (T0313)D157 (T0313)V164 (T0313)Q161 hbond (T0313)C155 1 try4-opt2.sheets: SheetConstraint (T0313)L154 (T0313)D157 (T0313)V164 (T0313)Q161 hbond (T0313)C155 1 The servers seem to be hedging their bets on this strand alignment. I'm too tired to make a reasonable decision, so I'm just going to polish up the try* models and submit 5 different ones. The cores are all pretty similar, and I don't think we'll get these other alignments any better than they are now. order of tries with different costfcns: xntry1.costfcn try1, try3, try2, try4 try2.costfcn try3, try4, try2, try1 try3.costfcn try3, try4, try2, try1 try4.costfcn try3, try4, try2, try1 unconstrained.costfcn try3, try4, try1, try2 Tue Jun 20 16:02:30 PDT 2006 Kevin Karplus Starting try5 (from existing models, including robetta models) on camano. Tue Jun 20 21:53:19 PDT 2006 Kevin Karplus try5 polished up try3-opt2 With try5.costfcn try5, try3, robetta5, try2, try4, robetta3, try1 Rosetta also likes repacking try5 best (wasn't compared with robetta models). I think I'll submit try5-opt2 # best-scoring, but D31 squeezed in a bit try4-opt2 from alignments without constraints try3-opt2 previous version of try5-opt2, strong constraints that failed try2-opt2 medium constraints that failed try1-opt2 fully automatic Tue Jun 20 22:11:36 PDT 2006 Kevin Karplus So submitted. Fri Jul 14 12:01:24 PDT 2006 Kevin Karplus Best server model: RAPTOR-ACE_TS4 -1.11 Our best model: try4-opt2.gromacs0 -0.96 best submitted: model2 -0.96 our best server: SAM_T06_server_TS1 -0.90 14th of 54 TS1