Kevin Karplus 10 May 1998 t51 is SW:GLME_CLOCO (EC 5.4.99.1) t51 has no close homologs found in a target98 alignment (the reported homology is really for t50 in the same crystal, not t51). The highest scoring sequence with the target model is 1aei[ABCDEF] at -3.880 (fssp representative 1avc) The highest scoring template model is 1mrj at -2.790 Summing in both directions gives a boost to 1svb, making it best at -4.04 delta true delta false approx new fold prob domains 67 479 88% domains 9 110 92% fssp-test 18.02 100 85% We may need to look for some functional similarity before venturing to make any of these weak and inconsistent predictions. -------------------------------------------------- Kevin Karplus 10 May 1998 Note: homologies for t50 may help as the known homologies there are to parts of much larger proteins, and t51 may have similarity to a different part. Unfortunately, none of the high scores there are particularly high here. 23 July 1998 The fold representative for 1mdl is 2mnr. Way down on the list of related structures is 1reqB with a zscore of 8.2 over 195 residues. Since this structure is not in our library, it may well be worth investigating since it's function is that of a methylmalonyl-coa mutase (E.C. 5.4.99.2) while t51 (GLME_CLOCO) is a METHYLASPARTATE MUTASE (E.C. 5.4.99.1). I'm sure why 1req doesn't exist in the rdb score files for this target, given that it is in the library and target98.ids. 25 July 1998 Kevin Karplus Moved all the old results to "old" and am re-running with the newest version of the Makefile. I'll also try concatentating the unmatched tail of t50 onto the beginning of t51 and see if that finds anything. FOR T51: wu-blast: doesn't really find anything (1akhA 0.788) double-blast: doesn't find anything target model: very few homologs in t51.t98_6 best scores: 1svb (-5.66) 1a3g[ABC] (-4.14) 1aei[ABCDF] (-4.1) template model: best scores: 1mrj (-2.79) 1efgA_2 (-2.57) 1sesA (-2.33) summed: 1svb (-6.28) 1a3g[ABC] (-4.14) 1aei[ABCDF] (-4.1) t51.remote_4: 1ttt[ABC]=1tui[ABC] (-123.95) FOR T50TAIL+T51 (last 27 residues of T50, 10X, T51): wu-blast: doesn't really find anything (1akhA 0.875) double-blast: doesn't find anything target model: best scores: 1svb (-5.60) 1aei[ABCDF] (-4.4) 1a3g[ABC] (-4.05) template model: 1reqB (-6.05) 4hb1 (-5.18) 1aa7A (-4.3) 1dkgA (-4.24) summed: 1reqB (-6.05) 1svb (-6.05) 4hb1 (-5.18) 1etu (-4.46) 26 July 1998 Making the joint models for 1svb, 1reqB, 4hb1. Best non-self scores: 1reqB/1reqB-t50tail+t51-global t50tail+t51 520 -0.22 -13.21 1reqB/1reqB-t50tail+t51-post t50tail+t51 520 -1.61 -13.21 1reqB/1reqB-t51-global T0051 483 0.79 -12.92 1reqB/1reqB-t51-post T0051 483 -0.59 -12.92 1svb/t51-1svb-vit 1svb 395 -8.34 -9.82 4hb1/4hb1-t50tail+t51-vit t50tail+t51 520 -8.76 -8.78 1svb/t50tail+t51-1svb-vit 1svb 395 -8.33 -8.58 The 1reqB matches are only to the t51 part, not the t50 tail---why didn't they come up in the score-lib run for t51? Hmm---since t50 matches 1reqA fairly well, we might want to do a 1reqA, 1reqB match for t50 and t51. 28 July 1998 Unfortunately, the part of 1reqA that we match does NOT interact with 1reqB, except through a long ligand (coenzyme B12). The alignment 1reqB-t51-global is long, but has some nasty gaps. There are lots of other alignments that have similar residue identities. There are a lot of helices that can be lined up in different ways. The t51-1svb-global alignment is also quite choppy. 8 August 1998 Kevin Karplus I'm making the fssp and constrained alignments for 1reqB, in the hopes that these will give a better alignment. Interestingly, 1reqB has a high structural similarity to 1reqA (with only 24% residue identity). There are many similar structures with low residue id, so the constrained alignment is likely to be useful. (Note: at least some of the similarities are TIM barrels for both 1reqA and 1reqB, though they have more in common than just a TIM barrel, and the TIM barrels did not score t51 well.) Looking at the 1reqB, 1svb, 4hb1 alignment scores. The best alignments are 1reqB/1reqB-t51-global T0051 483 0.79 -12.92 1svb/t51-1svb-vit 1svb 395 -8.34 -9.82 4hb1/4hb1-t51-vit T0051 483 -7.19 -7.14 1svb/1svb-t51-vit T0051 483 -2.40 -2.38 1reqB/t51-1reqB-vit 1reqB 619 -2.64 -0.62 4hb1/t51-4hb1-vit 4hb1 44 -2.04 0.80 None of these works well in both directions, and 1svb is perhaps the least painful. I'm a bit suspicious of the 1reqB-t51-global score, since it requires reverse-sequence scoring to fix a bad raw score. 13 August 1998 I don't know WHAT I've been thinking---I knew that t50 was only one domain of 1reqA, but I never thought to check whether t51 matched 1reqA! Well, I checked, and it does quite well---here are the top non-self scores: 1reqA/1reqA-t51-const-global T0051 483 -18.78 -17.23 4hb1/4hb1-t51-global 4hb1 44 -37.18 -14.61 4hb1/4hb1-t51-post 4hb1 44 -38.56 -14.61 1reqB/1reqB-t51-global T0051 483 0.79 -12.92 1reqB/1reqB-t51-post T0051 483 -0.59 -12.92 1svb/t51-1svb-vit 1svb 395 -8.34 -9.82 4hb1/4hb1-t51-vit T0051 483 -7.19 -7.14 1reqA/1reqA-t51-vit T0051 483 -4.54 -3.99 1reqB/1reqB-t51-vit T0051 483 -5.12 -3.78 4hb1/4hb1-t51-global T0051 483 -7.19 -2.76 4hb1/4hb1-t51-post T0051 483 -8.57 -2.76 1svb/1svb-t51-vit T0051 483 -2.40 -2.38 1reqB/t51-1reqB-vit 1reqB 619 -2.64 -0.62 1reqA/t51-1reqA-vit 1reqA 727 -0.14 -0.56 1reqB/t51-1reqB-global 1reqB 619 -3.13 -0.34 1reqB/t51-1reqB-post 1reqB 619 -4.52 -0.34 Since t50 matches at the END of 1reqA, I'll try a chimeric sequence: t51+t50. (I tried t50tail+t51 before, but that was clearly the wrong way around.) The alignments with 1reqA look moderately promising---I'll have to look at them in 3D. I may want to try again with t51+50X+t50. 15 August 1998 I replaced the Makefile, and redid the fssp alignments to use just fssp and not fssp-t98. It may be worth examining them again (particularly the t51+t50 fssp alignment), perhaps hand-aligning again. The best-scoring fssp alignments (viterbi scoring) from the library are 1idy -8.420 1svb -6.650 1ovaA -6.470 1mrj -5.690 1oacA -5.630 1hrdA -5.620 1iba -4.750 2abk -4.570 The best scoring t98 alignments from the library (viterbi scoring) are 1rcd -6.820 1efgA_2 -6.200 1mseC_1 -6.180 1ad2 -5.160 1mrj -4.840 1idy -4.320 1jud -4.040 2cae -4.040 1cae -3.900 1hrdA -3.810 1tlfA -3.000 Mon Aug 17 11:10:54 PDT 1998 The best non-self alignment score is still 1reqA/1reqA-t51-const-global T0051 483 -18.78 -17.23 I looked at the alignments again, and my favorite alignment is now t51+t50/1reqA/1reqA-t51+t50-hand4.a2m I'm not sure I'm going to get any more out of further computation or viewing, so maybe it is time to submit this. I'll trim out the t50 alignment to get 1reqA/1reqA-t51-hand4.a2m OOPS--I never looked at 1bmtA, what an oversight! Hmm---it scores well, but 1bmtA is much shorter than t51+t50, and the only way a decent alignment is possible is if t51+t50 covers two copies of 1bmtA (the dimer). Attempts to align t51+t50 with 1bmtA_20X_1bmtB were not very succeessful so I think I'll stick with the 1reqA prediction.