9 June 1998 Kevin Karplus Despite the answer to the question about homologous sequences, t62 has a very strong homology to 1cnf, 2cnd, 1cne, and 1ndh. (FSSP rep 1ndh) These are reductases also, and the next several matches are structurally similar reductases (1fnc, 2pia, and 1fdr as FSSP reps). Our best bet is probably to use 1ndh.fssp as a base for the alignment, but we may want to align to 1cnf or 2cnd using it. The model library alignments like 2cnd_2 and 2pia_2 best---the second domain of these reductases, though the first domains score well enough that one should probably look for a complete alignment. Somewhat surprisingly, the 1ndh alignment scores so badly it doesn't even appear on the list! It looks like 2cnd is our strongest candidate for alignment. It may be worth extracting a subtree from the rather larger target98_6 alignment---grabbing just enough of the other subfamilies to get the closest 3D structures (probalby 2cnd and 2pia). 10 June 1998 Kevin Karplus If we split the t62.t98_6.tree into its two main subfamilies, 2pia comes in the same subtree as t0062, but 1cnd and 1ndh occur in the other subtree. Note: all these are the same SCOP superfamily Ferredoxin reductase-like, C-terminal NADP-linked domain, but there is a split between the reductases and Phthalate dioxygenase reductase (2pia). Perhaps we should start with a DALI alignment of 2cnd, 1ndh, and 2pia as a seed for contrained model building? -------------------------------------------------- From compbio.casp-request Wed Jun 10 11:44:53 1998 Return-Path: karplus@cse.ucsc.edu Date: Wed, 10 Jun 1998 11:43:34 -0700 From: Kevin Karplus To: Margareta@xray.bmc.uu.se Cc: compbio.casp@cse.ucsc.edu Subject: Flavin reductase for CASP contest Thank you for submitting Flavin reductase as target 62 for the CASP-3 prediction contest. I was wondering about the answer to question 7 "Homologous Sequence of known structure? no". A simple blast search of PDB gets Smallest Sum High Probability Sequences producing High-scoring Segment Pairs: Score P(N) N pdb|1CNE| Nitrate Reductase (Cytochrome B Reductase Fra... 65 0.0068 2 pdb|1CNF| Nitrate Reductase (Cytochrome B Reductase Fra... 65 0.0068 2 pdb|1CNV| Crystal Structure Of Concanavalin B At 1.65 A... 55 0.55 1 pdb|1NDH| Cytochrome B5 Reductase (E.C.1.6.2.2) 53 0.78 1 and the 1CNE and 1CNF scores look strong enough to claim some homology, even if BLAST only aligns 54 of the 232 residues. Perhaps you might want to add a comment about this homology to your submission, clarifying your answer to question 7. Note: I would prefer not to get any information from you EXCEPT through the official CASP-3 web pages, so that my predictions can't be "tainted" by extra information. -------------------------------------------------- 10 June 1998 Kevin Karplus The alignment t62-2cnd-global.pw looks excellent. The alignment 1ndh-t62-fssp-global.pw looks ok, but the residue identities are very few---it might be better to align 2cnd using the 1ndh fssp model. From compbio.casp-request Thu Jun 11 00:52:59 1998 Return-Path: mi@alpha2.bmc.uu.se Date: Thu, 11 Jun 1998 09:52:01 +0200 (MET DST) From: Margareta Ingelman To: Kevin Karplus Cc: Margareta@xray.bmc.uu.se, compbio.casp@cse.ucsc.edu Subject: Re: Flavin reductase for CASP contest In-Reply-To: <199806101843.LAA23954@purr.cse.ucsc.edu> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Hallo, Sorry for the somewhat incomplete answer to the question about homology. Yes - there is some homology to the proteins of the ferredoxin reductase family but quite low just based on a sequence comparison. Best regards, Margareta Ingelman Margareta Ingelman Department of Molecular Biology Swedish University of Agricultural Sciences Uppsala Biomedical Centre P.O. Box 590 S-751 24 Uppsala, Sweden Phone: +46-18-17 49 00, Fax: +46-18-53 69 71 e-mail: margareta@xray.bmc.uu.se or margareta.ingelman@molbio.slu.se or mi@alpha2.bmc.uu.se On Wed, 10 Jun 1998, Kevin Karplus wrote: > > Thank you for submitting Flavin reductase as target 62 for the CASP-3 > prediction contest. I was wondering about the answer to question 7 > "Homologous Sequence of known structure? no". > > A simple blast search of PDB gets > Smallest > Sum > High Probability > Sequences producing High-scoring Segment Pairs: Score P(N) N > > pdb|1CNE| Nitrate Reductase (Cytochrome B Reductase Fra... 65 0.0068 2 > pdb|1CNF| Nitrate Reductase (Cytochrome B Reductase Fra... 65 0.0068 2 > pdb|1CNV| Crystal Structure Of Concanavalin B At 1.65 A... 55 0.55 1 > pdb|1NDH| Cytochrome B5 Reductase (E.C.1.6.2.2) 53 0.78 1 > > and the 1CNE and 1CNF scores look strong enough to claim some > homology, even if BLAST only aligns 54 of the 232 residues. > > Perhaps you might want to add a comment about this homology to your > submission, clarifying your answer to question 7. > > Note: I would prefer not to get any information from you EXCEPT > through the official CASP-3 web pages, so that my predictions can't be > "tainted" by extra information. 11 June 1998 Kevin Karplus While it is practically certain that T0062 should be modeled by one of the reductases, we have to decide which of the three SCOP families we should pick. There are two methods to try---what scores best and which is closest in the eveolutionary tree. Since the scoring may be heavily dependent on how many homologs there are and how diverse a set they are, the evolutionary tree approach seems a bit more robust. On the tree for t62.t98_6, 2pia was the closest to the target, but not all the reductases were represented in the tree. Also, the alignment may not be the best possible alignment of reductases, since no structural information was used. I'm building an evolutionary tree of the reductases using the alignment 1ndh.constr-t98.a2m, which I believe is the best alignment of them that we have (though we may want to try from 2pia.constr-t98.a2m also). T0062 is SW:UBIB_ECOLI in NRP, so I'll look for that in the resulting tree---again it seems to be much closer to 2pia (Phthalate dioxygenase reductase E.C.1.18.1.-) than to the other reductases. I built a tree from the much larger 2pia.constr-t98.a2m alignment, and again it looks like 2pia is our closest 3d model. We'll have to look at the alignments t62-1ndh, t62-2cnd-global, and t62-2pia (currently the best joint models) and 1ndh-t62-fssp-global and 2pia-t62-const-global (currently the best models from structures). The 2pia models may be scoring lower because of all the ferredoxins mixed into the 2pia alignment. It may be worth making a smaller alignment with just the smallest subtree containing UBIB and 2PIA, and using that for alignment. Note: although the 1ndh-t62-fssp-global alignment scores t62 much better than 2pia-t62-const-global, the two const alignments score almost the same (-79.8, -77.37), so I'm now strongly inclined to pick the 2pia chain as the base model. 25 July 1998 Kevin Karplus Remaking with the newest Makefile (and using t62.t98_3 as the main alignment), to see if this makes any difference in whether 2pia or 1ndh is the better choice. wu-blast: 1cnd=1cnf=2cnd best at -7.156 double-blast: 1ndh -25.07, 1fdr -23.49, 2cnd -22.84, 1a8p -20.14, 2pia -15.33 26 July 1998 t62.t98_3: 1cnf=2cnd -149.19, 1cne -145.32, 1ndh -142.34, 2pia -75.04, template models: 2cnd_2 -73.210 2pia_2 -41.260 1fnb_2 -28.770 2pia_1 -27.060 2cnd_1 -11.970 sum: 1cnf=2cnd 1-49.19 1cne -145.32, 1ndh -142.34, 2pia -75.04 ... Hmm---maybe we should paste together the first part of 2pia with the second part of 2cnd? Why isn't the 1ndh model scoring well? Note: the NEW tree t62.t98_3.tree puts t0062 in the same subfamily as 1cnd and 1ndh now, NOT 2pia. Best non-self matches: 2cnd/t62-2cnd-global 2cnd 260 -158.32 -164.24 2cnd/t62-2cnd-post 2cnd 260 -159.71 -164.24 1ndh/t62-1ndh-global 1ndh 270 -151.73 -155.73 1ndh/t62-1ndh-post 1ndh 270 -153.12 -155.73 2cnd/t62-2cnd-vit 2cnd 260 -153.21 -151.18 1ndh/t62-1ndh-vit 1ndh 270 -146.98 -145.30 1ndh/1ndh-t62-fssp-global t0062 232 -95.54 -101.80 2pia/2pia-t62-vit t0062 232 -92.15 -90.68 2pia/2pia-t62-global t0062 232 -90.67 -89.19 2pia/2pia-t62-post t0062 232 -92.06 -89.19 2pia/t62-2pia-global 2pia 321 -87.23 -88.12 2pia/t62-2pia-post 2pia 321 -88.62 -88.12 1fdr/1fdr-t62-global t0062 232 -77.93 -83.67 1fdr/1fdr-t62-post t0062 232 -79.32 -83.67 1ndh/1ndh-t62-const-global t0062 232 -72.77 -78.92 2pia/2pia-t62-fssp-global t0062 232 -79.58 -77.71 2pia/2pia-t62-const-global t0062 232 -78.43 -77.21 1fdr/t62-1fdr-global 1fdr 244 -64.81 -68.96 1fdr/t62-1fdr-post 1fdr 244 -66.20 -68.96 2pia/t62-2pia-vit 2pia 321 -69.69 -68.80 t0062 TT--------LSCKVTSVEAIT-DTVYRVRIVPDAA--FSFRAGQYLMVV..MDE.RDKRPFSMAST L K S E I D P A F AG L V R S 2pia ..TTPQEDGFLRLKIASKEKIARDIWSFELTDPQGAPLPPFEAGANLTVA..VPN.GSRRTYSLCND t0062 TT--------LSCKVTSVEAITDTVYRVRIVPDAA...FSFRAGQYLMVV..MDE.RDKRPFSMAST C V R G V R S 2cnd GR--------IHCRLVAKKELSRDVRLFRFSLPSPdqvLGLPIGKHIFVCatIEGkLCMRAYTPTSM t0062 PDEKGFIELHIGAS.....EINL-YAKAVMDRILKDHQIVVDIPHGEAWL.....RDDEeRPMILIAGGT E D V P E L IL AGG 2pia SQERNRYVIAVKRD.....SNGRGGSISFIDDTSEGDAVEVSLPRNEFPL.....DKRA.KSFILVAGGI t0062 PDEKGFIELHIGAS.....EINL-YAKAVMDRILKDHQIVVDIPHGEAWL.....RDDEERPMILIAGGT DE G L L D I P G R I GG 2cnd VDEIGHFDLLVKVYfk9fpNGGL-MTQYL-DSLPVGSYIDVKGPLGHVEYt10ngKQRNARRLAMICGGS domain breaks: 2pia^ 2cnd^ t0062 GFSYARSILLTALAR..NPnRDITIYWGGREEQHLYDLCELEALsLKHP.GLQVVPVVEQ...P G S A R Y R EL P 2pia GITPMLSMARQLRAE..GL.RSFRLYYLTRDPEGTAFFDELTSD.EWRS.DVKIHHDHGD...P t0062 GFSYARSILLTALAR..NPNRDITIYWGGREEQHLYDLCELEALSLKHP.GLQVVPVVEQ...P G I L R E EL P L V V Q P 2cnd GITPMYQIIQAVLRDqpEDHTEMHLVYANRTEDDILLRDELDRWAAEYPdRLKVWYVIDQvkrP t0062 EaGWRGRTGTVLTAvlqdhgt..LAEHDIYIAGRFEMAKIARDLFcseRNAREDRLFGDAF Y G RD F 2pia T.KAFDFWSVFEKS.........KPAQHVYCCGPQALMDTVRDMT...GHWPSGTVHFESFgatntnar... t0062 EAGWRGRTGTVLTAVLQDHGT..LAEHDIYIAGRFEMAKIARDLFCSERNAREDRLFGDAFAFI... E GW G VL AVL H G M A F 2cnd EEGWKYSVGFVTEAVLREHVPegGDDTLALACGPPPMIQFAISPNLEKMKYDMANSF-------vvf domain break: ^2pia It looks like 2pia for the first domain and 2cnd for the second domain gives the best match. 27 July 1998 Kevin Karplus Took the alignments above, reduced them to the relevant domains and tweaked to get 2pia_1-hand and 2cnd_2-hand. 2pia_1-hand has 23 conserved residues, 3 gaps (1,2,2 residues) and 1 insert (1 residue). By fiddling with 2cnd_1, I can get it up to 19 conserved residues, with 4 gaps (3,2,1,9) and 2 insertions (1,1). Not bad, but not as good as 2pia_1-hand. 2cnd_2-hand.a2m has 31 conserved residues and 4 gaps (2,1,3,2) all on exposed loops. The t62.t98_6.2d prediction looks a bit better than t62.t98_3.2d. I moved the boundary between the two predictions to a place were they looked fairly superimposable. From venc@september.llnl.gov Fri Jul 31 16:18:39 1998 Return-Path: venc@september.llnl.gov Date: Fri, 31 Jul 1998 16:17:41 -0700 From: venc@september.llnl.gov (Ceslovas Venclovas) To: karplus@cse.ucsc.edu Subject: Update for target T0062 Dear Predictor, It has been brought to our attention that there has been published discussion on functional and structural relationship for another CASP3 target (T0062). Please, see: Andrews SC; Shipley D; Keen JN; Findlay JB; Harrison PM; Guest JR. The haemoglobin-like protein (HMP) of Escherichia coli has ferrisiderophore reductase activity and its C-terminal domain shares homology with ferredoxin NADP+ reductases. FEBS Letters, 1992 , 302:247-52. Sincerely, Ceslovas Venclovas for CASP3 organizers -- Protein Structure Prediction Center, Lawrence Livermore National Laboratory, Livermore, CA 94550 E-mail: venclovas1@llnl.gov Phone: (925) 422-3097 Fax: (925) 423-3608