Wed Jun 23 08:48:55 PDT 2004 T0214 Due 18 Aug Wed Jun 23 11:18:48 PDT 2004 Kevin Karplus Probably fold-recognition b.43.* Wed Jun 23 16:09:45 PDT 2004 Kevin Karplus try1 is rather ugly---strands don't pair up much. The alignments used are generally quite short super-secondary alignments, and are not consistent---this may not be fold-recognition. The t2k secondary structure predictions seem to be fairly consistent. I've no idea what to do with this one yet, so let's just run try2 from all-align, with more super iterations but the same cost function. Fri Aug 6 14:00:26 PDT 2004 Kevin Karplus On the FORCASP web site (July-24th/04), Alexey Murzin wrote: Kill three birds with one stone. The targets T0227, T0213 and T0214 from three different sequence families probably belong to the same superfamily. They have similar sizes and predicted secondary structures, and, most importantly, there is a similar conserved sequence motif in the N-terminal region of each target. If the structure of one of these targets were known, it probably would be a Fold Recognition pick for the two others. Is there a common FR prediction for all three targets? My pick is the structure of conserved hypothetical protein from Enterococcus faecalis V583, 1T62. Incidentally, it has been released by PDB just one day before the announcement of T0213 and T0214. 1T62 is a new member of the PUA-like fold. Its specific feature, a helix-rich insertion between the two last strands of the PUA-like fold, flanks the putative active site, that, in the crystal structure, binds the C-terminal His-tag of adjacent molecule. The conserved residues of each target map readily on this site. The modeling of other regions may be improved by using fragments of other PUA-like fold structures, whereas a variable helix-rich insertion may require a bit of 'ab initio' modeling. 1t62A was in the template library when T0227 was scored, but did not score all that well. Perhaps searches should be redone for T0213 and T0214. Fri Aug 6 19:36:25 PDT 2004 Kevin Karplus 1t62A does NOT score particularly well for T0214. How did Alexey conclude that T0213, T0214, and T0227 were from the same superfamily? Is the common motif really enough? I'll try making a multiple alignment of all three, with T0214 as the initial seed. The work will be done in subdirectory 14+13+27, using T0214 as the focus of the alignment. Fri Aug 6 20:39:18 PDT 2004 Kevin Karplus The first attempt to align T0214,T0213, and T0227, using the "realign_a2m" script does not have a very good result. The N terminus aligns fairly weill, but there is an EKH motif in T0213 and T0227 that is off by 1. (align1.a2m) Aligning them to the T0214.w0.5 HMM has even worse results (align2.a2m). Using just muscle without buildmodel has the same effect as using both (align3.a2m=align1.a2m). Using just buildmodel without muscle aligns only up to residue 40, then inserts long gaps (align4.a2m). I did a hand alignment which was similar to align1 and align3, but made one T in T0213 be an insertion, so that the EK[HE] motif lined up. Sat Aug 7 11:35:16 PDT 2004 Kevin Karplus In 14+13+27, the best-scoring template for t2k is 1t62A (probably because T0213 matches it fairly well). For t04, it comes out second, after 1r13A. The t2k alignment to 1t62A is lining up the beta sheets, but not forming the Hbonds. The stripes across the beta sheets at the two ends are R50-E106-K21 and R55-V101-L26. SheetConstraint R50 R55 E106 V101 SheetConstraint K21 L26 E106 V101 There is also a hairpin that aligns I40 with L47 (but this means an improbable pairing of F42 with G45). A better pairing might be F42 with L47, which would put E43 with K46: SheetConstraint V39 E43 R50 K46 Hbond E43 ---------------------------------------- I'll redo the make to get t04.many.frag fragments, and run try3 using both the alignments from this directory and from the subdirectory 14+13+27. The robetta models currently score better with the try3 costfcn than do the try1 or try2 models. I'll also pick up all the top hits (even the bad ones) from t2k and t04 in both this directory and 14+13+27 and add them as extra alignments. Sun Aug 8 10:11:47 PDT 2004 Kevin Karplus On try3, the results look pretty good, getting the same fold as for T0213. try3-opt2 outscores the robetta models (unlike try1 and try2) on the try3.costfcn. Wed Aug 11 17:46:26 PDT 2004 Kevin Karplus try4-opt2 looks fairly good, and scores a bit better than try3-opt2. I made an unconstrained.costfcn, which scores try4-opt2 best. Rosetta likes try1-opt2.repack-nonPC best. From andypohl@gmail.com Wed Aug 11 18:38:30 2004 MIME-Version: 1.0 Date: Wed, 11 Aug 2004 18:38:29 -0700 From: andy pohl To: Kevin Karplus Subject: sorry aleksey logged out of abyss and clicked reboot or shutdown instead of log out. we noticed a job you may or may not have finished running on abyss (it was slow). in any case, aleksey is leaving so the computer will be available now. sorry for the inconvenience. andy pohl Thu Aug 12 12:13:22 PDT 2004 Kevin Karplus I see a problem that I missed before in both try4 and try5. There is a large hydrophobic cleft that could be closed up if we pull E67 and K20 out of the way. Maybe we should request a salt bridge? (hbonds K20.NZ and E67.OE1 and OE2) Maybe a contact constraint for I22.CD1 and L64.CD2? and G75.CA-F11.CD1 Thu Aug 12 16:43:37 PDT 2004 Kevin Karplus try6 does not close the cleft. It looks like closing it would require a hinging motion at S59 and G99, which is difficult for undertaker to do. Should I try, or should I give up? I'll submit T0214.try6-opt2 # best score T0214.try6-opt2.repack-nonPC # best Rosetta T0214.try1-opt2 # full auto 14+13+27/decoys/T0214.try1-opt2 # auto from multiple align T0214-1jnyA-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 Thu Nov 18 23:05:14 PST 2004 Martina Koeva Based on the smooth gdt scores: best sam-t04 36.5589 (try5-opt1, which we did not submit) best submit 36.1787 (model2) model1 36.1742 auto 22.3709 align 20.8267 robetta best 28.4129 (robetta model9) robetta1 17.6781