Mon Aug 5 19:18:44 PDT 2002 t0165 Clear homology to 1a88A c.69.1.12 There may be a separate domain for first 50-60 residues. 7 Aug 2002 Kevin Karplus In try1-opt-scwrl, the first sheet looks pretty good (needs a few constraints to keep the last strand from drifting, and needs several constraints to reattach the first strand). Look at the alignments to see if we can get the first strand Hbonds right. 21 Aug 2002 Kevin Karplus Increased weight of dry6.5 and dry8 terms, and rescored. try1-opt-scwrl still best. Added try2.constraints to keep the strand from drifting, but try1-opt-scwrl still best. Thu Aug 22 13:39:14 PDT 2002 Kevin Karplus It looks like V56-K62 should be antiparallel to R68-V75 Let's try adding CB constraints for 55> kVYRLTYK sfgha 76< pVAYWGTIR Doing so still leaves try2-opt as the best score. Let's try a subdomain t0165-1-60, and do a full search there, then start try3 with extra alignments from that subdomain. Fri Aug 23 10:36:01 PDT 2002 Kevin Karplus New best score is try3-opt. There is a tremendously bad break between A84 and I85 and between A254 and D255. Upping the break penalty from 5 to 15 makes try2-opt look best. Perhaps I should do a short polishing run with try2-opt and try3-opt as initial conformations, then do a VAST run to see if I can get a better alignment to work from. Fri Aug 23 16:00:54 PDT 2002 Kevin Karplus Best score now try4-opt, but I really don't like the detachment of K55-Y61 from the sheet. V216-D249 looks like is is out in space also. I like try3 much better than try4 or try2, despite the terrible break. Let's try sending try3-opt to vast: run VS30774 password casp5t0165 21:47 The VAST job is STILL running! 22:21 STILL!! 24 Aug 2002 Kevin Karplus Best hit in vast was 1c4xA (1.7 rmsd over 223 residues). Other high scores are 1a88A (2.9 over 196), 1a8s (2.9 over 193), 1cqwA (2.9 over 204), 1qtrA (3.4 over 220), 1bn7A (2.8 over 201), ... I'll start a new run with just one initial alignment, a trimmed-back version of the T0165-1c4xA alignment. When that finishes, if it isn't best, I'll do a polishing run with it and the current best. 16:27 24 Aug 2002 Best score is still try4-try2.0.60, with try4-opt close behind. I'll do a polishing run with those two and with try5-opt. 24 Aug 2002 I think that try3 still looks much better than try4, try5, or try6, all of which blow up the beta sheet. I'm going to have to start over from try3-opt. First, break off V75-A84, so that they can be repositioned to get rid of the A84-I85 gap. Also move L245-A254, to try to get some stretchiness to close the A254-D255 gap. I'd better add some hbond constraints to keep the sheet together. Copying all the sheet Hbonds from try3-opt gives me try7.constraints. With these constraints, naturally try3-opt scores best. Even better would be to copy the Hbonds from T0165-1cx4A.pdb, from the VAST alignment. These are consistent with the try3-opt hbonds, but extend some strands slightly, so I added them to try7.constraints. Now try3-al10.5.80 scores best. Now I need to redo the undertaker run that started from the T0165-1cx4A alignment, with the Hbond constraints in place to keep from blowing up the sheet. I'll start a 12-hour run and look at it in the morning. With any luck, it'll only need an hour's polishing. 25 Aug 2002 7:51 try7-opt IS more compact that previous runs, and scores better, but seems to have unhinged the region from about N209-M251. I think I need some constraints for L245-L135 T237-A189. Or maybe not. In T0165-1cx4A, the pair of helices flops over to cover the end of the sheet, with T237 coming close to G119, and L245 coming close to Q182. It is difficult to figure out how this pair of helices flops, because of the very polar nature of R230-K243. The non-polar face A241,T244L245 needs to be buried though. Perhaps rather than hinging, what is needed is to slide G233-F248 down into the groove, with L245 near I188 and T237 near S181. I'd have to move pieces to allow rearrangement, moving G233-V259 and D203-R230, and trying to reassemble the bits. I'd probably want to move 1-17 out of the way also. 10:57 try8-opt is new best, but it has not managed to press the helix S234-L245 down into the groove. Increasing the constraint weight, decreasing break weight, and adding a new constraint to try to bring S201 and A84 together makes one of the try3 runs score best. Let's try starting with try8-opt-scwrl, and doing TryAllAlign on top of it with the new scoring function. 12:47 The TryAllAlign did no good, but the ordinary optimization made try9-opt the new best. There seems to be a knot with Y247-A254 on the wrong side of K198. Let's move it manually and reoptimize 15:04 try10-opt is new best and the knot is gone, but it needs more gap closure. Let's try to polish it, then submit. Although try11-opt is current best, I don't like the break at P260-V261, and the S227-F228 gap is terrible. The break penalty must be set WAY too low. Raising it from 5 to 15 still leaves try11 on top. Raising it up to 25 may be a bit extreme, but it will guide another polishing run, which I'll start now. 25 Aug 2002 21:23 try12.0.70 is new best, beating out try12.1.70 and try12-opt slightly. If we drop the break penalty from 25 to 20 and raise clash penalty from 1.5 to 2, try12.0.70 is still the best. There are some breaks and clashes, but the packing is pretty good. Although we could probably improve it further, I'm going to declare this "good enough" and give up.