Fri Jul 11 09:14:21 PDT 2008 T0499 Make started Fri Jul 11 09:15:23 PDT 2008 Running on peep.cse.ucsc.edu Fri Jul 11 11:30:25 PDT 2008 Kevin Karplus This target is very similar to T0499. I don't see the point to human prediction for NMR models that have over 60% identity t o lots of existing templates. Fri Jul 11 11:41:11 PDT 2008 Kevin Karplus For try2, I'll try including a lot of the templates that aren't in the template library, and see if that makes for any difference in the model. I'll also increase breaks and clashes substantially. When that is done, I'll do one polishing run as try3 and submit---this target doesn't seem to be worth much of my time. Fri Jul 11 14:51:43 PDT 2008 Kevin Karplus The try2-opt3 model looks good and scores well. Rosetta likes the try2-opt3.gromacs0.repack-nonPC model. Fri Jul 11 17:18:31 PDT 2008 Kevin Karplus OK, try3-opt3 and try2-opt3.gromacs0.repack-nonPC are about as good as I'm going to get, so this is done until the metaserver results come in. Tue Jul 15 11:26:49 PDT 2008 Kevin Karplus MQAU likes SAM-T08-server models, pipe_int_TS1, SAM-T06-server models, pro-sp3-TASSER_TS1, all with predicted GDT around 81% MQAC likes Phyre_de_novo_TS1 0.887 MULTICOM-CLUSTER_TS1 0.887 MULTICOM-RANK_TS3 0.887 MULTICOM-CLUSTER_TS4 0.886 SAM-T08-server_TS1 0.885 MULTICOM-RANK_TS2 0.885 SAM-T06-server_TS4 0.885 Pcons_multi_TS5 0.885 ... I've started MQAC1 and MQAU1 metaserver runs, but I'll also start and MQAX3 run from all servers *except* SAM servers. Tue Jul 15 11:31:17 PDT 2008 Kevin Karplus Both MQAC1 and MQAU1 are favoring the SAM-T08-server_TS1 model. Tue Jul 15 12:15:58 PDT 2008 Kevin Karplus MQAX3 is favoring Pcons_multi_TS5 (and maybe YASARA_TS5). Tue Jul 15 12:55:42 PDT 2008 Kevin Karplus costfcn favorite models try1 MQAC1-opt3 MAU1-opt3 try3-opt3 try1-opt3 try2-opt3 MQAX3-opt3 try2 try3-opt3 MQAX3-opt3 MQAC1-opt3.gromacs0 try2-opt3 try1-opt3.gromacs0 MQAU1-opt3.gromacs0 try3 MQAX3-opt3 try3-opt3 MQAC1-opt3.gromacs0 try1-opt3.gromacs0 MQAU1-opt3.gromacs0 try2-opt3 rosetta try2 try3 MQAC1 MQAU1 try1 MQAX3 (all .gromacs0.repack-nonPC) Tue Jul 15 13:40:12 PDT 2008 Kevin Karplus The backbones of the models are almost identical, but MQAC1 swings the sidechain for W43 out of the core, and rearranges other aromatics (F52 and Y45) to accomodate this change. I believe that this incorrect, so I'm moving MQAC1 to the end of my list. Since this is an NMR model, and solvent-exposed polar sidechains are not likely to have many restraints, I should model them in as similar a was as possible to the NMR modeling techinques. Rosetta's favorite sidechains are more canonical them undertaker's, so I should probably use the rosetta-optimized versions of the models. This reduces the variety of the sidechains, and (somewhat surprisingly) Rosetta actually sticks W43 back into the core for MQAC1. I like try3-opt3.gromacs0.repack-nonPC best, as it has the ring of Y45 properly buried, and its OH atom close to Y3.OH (4.97 Ang). Maybe I should do one more run of undertaker, with those Y.OH atoms constrained, and W43.CZ3 within 3.6 (ideally 3.42) of L5.CD1, to select models that keep the aromatics in the core. Tue Jul 15 14:16:56 PDT 2008 Kevin Karplus The template 2igdA has the corresponding W buried and the two TYR.OH atoms 4.86 Ang apart, so I believe that forcing these properties will help the models. Maybe I should look again at the T0498 models also, to make sure they don't have similar problems. Tue Jul 15 14:23:33 PDT 2008 Kevin Karplus I'm also starting an MQAX4 run to see if there is a metaserver model that packs the Y45 residue well. Tue Jul 15 15:28:24 PDT 2008 Kevin Karplus The MQAX4 run is based on pipe_int_TS1. I like try4-opt3 best (better than the gromacs0.repack-nonPC version of it, because rosetta kicks Y29 out into solution while undertaker had the aromatic ring packed up against K18 and I33. The lysines in the repacked version are more likely to be the ones chosen by the NMR modeling program, though. All the models look a little foamy to me, but they are all so close that it is hard to justify doing anything to them. Tue Jul 15 16:02:58 PDT 2008 Kevin Karplus Submitted with comment: All the models, whether native SAM+undertaker models or metaserver models, were very similar. I added some constraints to keep the OH atoms of Y3 and Y45 close together (as seen in the template 2igdA) and to keep W43 buried. The first model is the best scoring with my constrained cost function. The second is the same model followed by energy minimization by gromacs and repacking the sidechains by rosetta. Although many of the solvent-exposed residues have been changed to more canonical rotamers, I don't really like the way rosetta moved Y29, so I ranked this model lower. The third model is a meta-server model, optimized by undertaker using the same constrained cost function used to generate the first model, but starting from non-SAM server models. The fourth model is from an earlier metaserver run, without the constraints on Y45, so it ended up kicked out of position. The fifth model is the fully automatic one, with no human intervention other than choosing the gromacs-minimized version of the model to minimize clashes. Model 1 T0499.try4-opt3.pdb # < try3-gromacs0.repack-nonPC < try2-opt3 < align(2igdA) 2 T0499.try4-opt3.gromacs0.repack-nonPC.pdb # < try3-gromacs0.repack-nonPC < try2-opt3 < align(2igdA) # best rosetta energy 3 T0499.MQAX4-opt3.pdb # < pipe_int_TS1 4 T0499.MQAC1-opt3.gromacs0.repack-nonPC.pdb # < SAM-T08-server_TS1 5 T0499.try1-opt3.gromacs0.pdb # < align(2gi9A) Thu Jul 17 17:16:23 PDT 2008 SAM-T08-MQAO hand QA T0499 Submitted Thu Jul 17 17:16:24 PDT 2008 SAM-T08-MQAU hand QA T0499 Submitted Thu Jul 17 17:16:24 PDT 2008 SAM-T08-MQAC hand QA T0499 Submitted