Wed Jul 9 09:55:31 PDT 2008 T0493 Make started Wed Jul 9 09:55:59 PDT 2008 Running on cheep.cse.ucsc.edu Wed Jul 9 13:40:28 PDT 2008 Kevin Karplus This looks like a c.37.1.19 family member, with all the labeled hits out to E-value 4 being from that family, and with hits as good as 8e-20. Wed Jul 9 15:11:33 PDT 2008 Kevin Karplus All the models (try1 and alignments) are very close. There are some gaps to close, but the sheets and helices look good. Wed Jul 9 17:00:28 PDT 2008 Kevin Karplus try2-opt3 cuts the breaks and clashes in half, but doesn't eliminate them. gromacs does not magange to make further reductions in either. I may have to move breaks WAY up and increase the application of ReduceBreak to try to rearrange the loop. If that doesn't work, I may hve to manually move the pieces and let it reassemble. Wed Jul 9 19:40:42 PDT 2008 Kevin Karplus Breaks and clashes were cut way down in try3-opt3 but gromacs manages to decrease them further, with only small damage to other components of the cost function. I should probably do another polishing run, with breaks and clashes higher still. I'd also like to try to squeeze V88 and Q96 closer together---maybe as close as anti-parallel bonded residues (CA separation of 5.25, CB sep of 5.02). For try4, I'll turn the breaks and clashes down i norder to try to do the squeeze. If that doesn't work, I might want to try running try4 from alignments, where pieces may be more mobile. Wed Jul 9 22:03:40 PDT 2008 Kevin Karplus try4-opt3 closed gaps nicely, but didn't attempt the squeeze. Rosetta likes try4-opt3.gromacs0.repack-nonPC fairly well. For try5, I'll use the same costfcn as for try4, but starting from alignments, rather than models that have already had some gap closure. Thu Jul 10 12:45:47 PDT 2008 Kevin Karplus Try5 scores ok, though not quite as good as try2. Try5 has not succeeded in closing gaps, but in many ways I like it better than the try1-try4 series. It has essentially the same underlying alignment. I'll try polishing it for try6. Sun Jul 13 07:50:06 PDT 2008 Kevin Karplus The MQAC quality assessment liked the MULTICOM-REFINE models best. The SAM_T08-server models are far down in the ranks (rank 82), but only slightly lower in predicted GDT. The MQAU quality assesment also likes the MULTICOM-REFINE models, but scores SAM-T08-server_TS2 fairly high (rank 11). I've started MQAU1 and MQAC1 runs, also MQAU4 and MQAC4. I may want to use the try6 costfcn also. Sun Jul 13 08:01:57 PDT 2008 Kevin Karplus I set up MQAC6 and MQAU6 also, though I imagine that the metaserver runs will probably all end up polishing the same one or two models. Sun Jul 13 10:49:43 PDT 2008 Kevin Karplus costfcn top models try6 try4-opt3 MQAU6-opt3 MQAC4-opt3 MQAU4-opt3 MQAC6-opt3 MQAC1-opt3.gromacs0 MQAU1-opt3.gromacs0 try6-opt3 try4 try4-opt3 MQAC4-opt3 MQAU4-opt3 try3-opt3 MQAU6-opt3 MQAC1-opt3.gromacs0 MQAU1-opt3.gromacs0 try6-opt3 MQAC6-opt3 try1 MQAU1-opt3 MQAC1-opt3 try4-opt3 MQAU6-opt3 MQAU6-opt3 MQAU4-opt3 MQAc^-opt3 try3-opt3.gromacs0 rosetta MQAC4-opt3.gromacs0.repack-nonPC MQAU4-opt3.gromacs0.repack-nonPC MQAC6-opt3.gromacs0.repack-nonPC MQAU1-opt3.gromacs0.repack-nonPC Sun Jul 13 12:52:14 PDT 2008 Kevin Karplus I now have 5 distinct models: try4, MQAU6, MQAU4, try6, try1. The main differences are the N and C termini and the loop from Q92 to V102. I don't know what to do about the loop, but for the termini, I rather like having a little beta-strand at each end. At the N-terminus that would be MQAU4 At the C-terminus that would be try4, MQAU6, or try1. To get both I can patch a little of MQAU4 into try4, say S1-Q9. Sun Jul 13 13:05:16 PDT 2008 Kevin Karplus try7 will try to optimize chimera-MQAU4-try4 Actually, I like the MQAU6 C-terminus a little better than the try4 one, so when try7 is done, I'll try patching in approximately L167-K174 and doing another optimization. Sun Jul 13 13:57:56 PDT 2008 Kevin Karplus try7-opt3 is the best with the try7 costfcn. MQAC4-opt3.gromacs0.repack-nonPC is the best with the Rosetta energy function. Sun Jul 13 14:11:52 PDT 2008 Kevin Karplus I started try8 to optimize chimera-try7-MQAU6. This shuld take care of the two termini, but what should I do about the I91-G101 loop, for which I have many different positions? I like the try7 loop, since it buries L98 and does the best job of covering the residues predicted to be highly buried. Sun Jul 13 15:03:57 PDT 2008 Kevin Karplus try8-opt3 scores best with the try8 costfcn, but Rosetta still prefers MQAC4-opt3.gromacs0.repack-nonPC. Sun Jul 13 15:18:37 PDT 2008 Kevin Karplus Submitted with comment: The first model is based mainly on SAM+undertaker models, but little bits of the N and C termini were copied from metaserver models. The I91-G101 loop is in quite different locations in different models. I like the version in Model 1, because it buries L98, but it is distinctly possible that this loop is a bit floppy in the real protein. The second and third models are meta-server models, optimized from server models. The 4th is another SAM+undertaker model, included to provide more variety. The 5th is the fully automatic SAM+undertaker model, without human intervention or metaserver work. Model 1 T0493.try8-opt3.pdb # < chimera-try7-MQAU6 chimera-try7-MQAU6 : mostly T0493.try7-opt3 < chimera-MQAU4-try4 L167-K174 from MQAU6-opt3 < MUProt_TS4 chimera-MQAU4-try4: S1-Q9 from MQAU4-opt3 < MULTICOM-CLUSTER_TS1 Q10-K174 T0493.try4-opt3.pdb try4-opt3 < try3-opt3.gromacs0