28 May 2000 Kevin Karplus T0094 (CPDase= cyclic phosphodiesterase) looks like another difficult target. Template model hits: chain E-value FSSP rep 2ts1_2 14 2ts1 1bncA 36 1bncA Target model hits: chain E-value FSSP rep 1ygh[AB] 3.5 1yghA 1fpv 25 2cas 1ijsP 28 2cas 2cas 29 2cas 4dpvZ 29 2cas blast hits chain E-value FSSP rep 1dml[BDFH] 5.8 1dmlB The predictions by three methods have apparently nothing in common: FSSP Z-scores: 2ts1 1bncA 1yghA 2cas 1dmlB 2ts1 54.3 5.3 -- -- -- 1bncA 67.5 -- -- -- 1yghA 34.6 -- -- 2cas 59.0 -- 1dmlB 7.6 The T2K alignment only found CPDase and one close homolog, so the target model can't be expected to generalize very far. (The T99 method, as recorded on the CAFASP server, didn't get anything better.) Perhaps we need to try doing a more speculative search for more remote homologs. Tue Jun 6 13:35:30 PDT 2000 Redid 2ry prediction with new neural net. Mon Jun 12 15:06:42 PDT 2000 Kevin Karplus Looked at the CAFASP predictions. Pretty strong consensus: num SCOP name 13 4.89.1 Acyl-CoA N-acyltransferases (Nat) 6 2.51.1 Core binding factor beta 4 3.3.1 FAD/NAD(P)-binding domain 4 2.1.1 Immunoglobulin 4 1.1.1 Globin-like Our predictions do include the most popular of these: 4.89.1. FSSP SCOP 2ts1 1.68.1, 3.19.1 1bncA 2.79.2, 3.25.1, 4.121.1 1yghA 4.89.1 2cas 2.9.1 1dmlB ? The alignment 1yghA/T0094-1yghA-global.pw looks pretty good. Mon Jun 26 09:52:14 PDT 2000 Remade 2ry predictions Wed Jul 26 17:08:37 PDT 2000 Running "make remote" changes the target model hits to e-value FSSP SCOP 1ygh[AB] 24. 1yghA 4.89.1.1.4 1bm9[AB] 34. 1bm9A 1.4.4.7.1 It still looks like 1yghA is our preferred template. 27 July 2000 Kevin Karplus Top hits with make 2track are: % ID Length Simple Reverse E-value SCOP 1dapA 320 -23.34 -9.83 1.1e-01 3.2.1.3.12, 4.66.1.2.1 2tbd 134 -20.48 -9.55 1.4e-01 4.74.1.1.1 1mla 309 -22.87 -8.31 4.9e-01 3.15.1.1.1, 4.48.20.1.1 1tc1A 220 -19.60 -8.31 4.9e-01 3.56.1.1.8 1tgoA 773 -19.61 -7.54 1.1e+00 3.50.3.6.4 1ba1 386 -19.39 -6.58 2.8e+00 3.50.1.1.1 1a2vA 655 -18.19 -6.43 3.2e+00 2.29.2.1.4, 4.15.2.1.4 1rypC 244 -16.50 -6.07 4.6e+00 4.132.1.4.4 1tyfA 193 -17.65 -5.94 5.2e+00 3.11.1.1.1 1gca 309 -18.97 -5.88 5.6e+00 3.88.1.1.5 1cg2A 393 -17.36 -5.85 5.7e+00 3.51.4.4.3, 4.48.18.1.1 1bk7A 190 -17.10 -5.73 6.5e+00 4.103.1.1.2 1oacA 727 -17.28 -5.53 7.9e+00 2.29.2.1.1, 4.15.2.1.1, 4.67.1.1.1 1qsmA 152 -17.90 -5.51 8.0e+00 4.89.1.1.6 1a3c 181 -18.70 -5.42 8.8e+00 3.56.1.1.12 1a0p 290 -15.65 -5.34 9.5e+00 1.60.9.1.2, 4.141.1.1.4 1bo1A 416 -15.95 -5.29 1.0e+01 4.122.1.2.1 1rtm1 149 -15.68 -5.15 1.1e+01 4.147.1.1.8, 8.1.1.1.1 1914 232 -15.72 -5.10 1.2e+01 4.41.1.1.1 2myo 118 -13.59 -5.00 1.3e+01 1.110.2.1.8 1sbp 310 -14.99 -4.95 1.4e+01 3.89.1.1.4 1dhn 121 -15.93 -4.88 1.5e+01 4.80.1.3.1 1a0aA 63 -15.86 -4.87 1.5e+01 1.40.1.1.5 These are all over the map---I think that the 2-track search is not working well, perhaps because it is using a poor null model for the 2ry structure track. 1qsmA comes up as another 4.89.1 hit, and that is the only hit that matches previous ones. Possible fold matches: 1mla 309 -22.87 -8.31 4.9e-01 3.15.1.1.1, 4.48.20.1.1 1cg2A 393 -17.36 -5.85 5.7e+00 3.51.4.4.3, 4.48.18.1.1 1tc1A 220 -19.60 -8.31 4.9e-01 3.56.1.1.8 1a3c 181 -18.70 -5.42 8.8e+00 3.56.1.1.12 1tgoA 773 -19.61 -7.54 1.1e+00 3.50.3.6.4 1ba1 386 -19.39 -6.58 2.8e+00 3.50.1.1.1 1a2vA 655 -18.19 -6.43 3.2e+00 2.29.2.1.4, 4.15.2.1.4 1oacA 727 -17.28 -5.53 7.9e+00 2.29.2.1.1, 4.15.2.1.1, 4.67.1.1.1 28 July 2000 Kevin Karplus I tried looking at the 2-track alignments. 1mla/T0094-1mla-2track-local needed to be edited to make the .stride.seq sequence align to the .stride-mixed.seq Resulting alignment has only 13 identical residues in 180 positions, and secondary structure alignment is not very impressive. 1tc1A/T0094-1tc1A-2track-local also needed to be edited for consistent alignment. 16 conserved residues. The first helix-strand-helix alignment looks good, but then a predicted strand is skipped in the middle of 1tc1A's second helix. Not very convincing. The pieces are pretty good 2ry matches, except residues HLSLLYA which we predict as helix, but align to a strand. 1tgoA/T0094-1tgoA-2track-local gets a nice 2ry structure match, but is not at all compact. 1a2vA/T0094-1a2vA-2track-local does not look too bad. It has good 2ry match except for one predicted helix matching a strand: HCKNH Residue conservation is reasonable: 21 residues. 1qsmA/T0094-1qsmA-2track-local looks terrible. Only 6 conserved residues and very poor 2ry structure match. 1yghA/T0094-1yghA-2track-local looks terrible. Only 6 conserved residues and very poor 2ry match. T0094-1dapA-2track-local only 14 conserved residues in a gapless alignment, and rather poor secondary structure match. Not very convincing. Looking at other alignments 2cas/T0094-2cas-local has a good short match to TVDRVSTGTFFFQCVFLLLQTT, which suggests that beta strand prediction may be a bit off, as TGTFF is part of a strand. Only useful as an undertaker fragment. 2ts1/2ts1-T0094-fssp-global has only 13 conserved residues in a very gappy alignment 1yghA/T0094-1yghA-global looks pretty good, with 33 conserved residues, but some of the 2ry structure prediction is way different. I edited the beginning slightly to get 1yghA/T0094-1yghA-global-edited. It still looks to me like 1yghA/T0094-1yghA-global-edited is our best bet, despite the 2ry structure disagreements. 30 July 2000 Kevin Karplus Remade 2track alignments using adptsyle 1, but T00994-1yghA-2track-local still looks bad to me. From casp4@bialko.llnl.gov Mon Jul 31 12:34:56 2000 Return-Path: Date: Mon, 31 Jul 2000 12:02:52 -0700 (PDT) From: casp4@bialko.llnl.gov (CASP4) To: sam-info@cse.ucsc.edu Subject: CASP4 T0094 target Dear CASP4 Predictor, Upon request of one of the predictors we have obtained more detailed information regarding His-tag for T0094 and updated the description of the target accordingly. Specifically, the following sentence was added to the 'Additional Information' section: "Due to the cloning strategy there are two additional residues and a six-fold his-tag (GSHHHHHH) fused to the C-terminus." CASP4 organizers 31 July 2000 Kevin Karplus Remade the 2track alignments again with new hmmscore. Top hits are still 1dapA, 2tbd, 1mla, ... though scores have changed. The 1yghA-T0094-1yghA-2track-local alignment looks good, but short. It is a gapless fragment of 1yghA/T0094-1yghA-global-edited.