Fri May 20 12:27:20 PDT 2005 Kevin Karplus miniAGRP is a short sequence taken from Loops and Links: Structural Insights into the Remarkable Function of the Agouti-Related Protein Glenn L. Millhauser, Joe C. McNulty, Pilgrim J. Jackson, Darren A. Thompson, Gregory S. Barsh, and Ira Gantz Annals of the New York Academy of Sciences 994:27-35 (2003) http://www.annalsnyas.org/cgi/content/full/994/1/27#F35 Glenn Millhauser's lab uses it for experimental work, as it is an antagonist for melanocortin receptors. The structure is already known (determined in Millhauser's lab), so the purpose of this bioinformatics study is to find out how conserved various residues are in the naturally occuring homologs. Fri May 20 15:07:31 PDT 2005 Kevin Karplus The protein "prediction" does not retain the disulfides---perhaps because putting the sidechains onto the templates does not pay attention to the original rotamer, even when the residues match. (This needs to be checked and possibly fixed.) I should be able to fix that damage by telling undertaker what disulfides I want, and it may get things right. I could also try to fix the damage by not running the initial fragments through SCWRL. Fri May 20 15:23:52 PDT 2005 Kevin Karplus Yep, it was SCWRLing the alignments that caused the disulfides to be lost. Including unSCWRLed versions of the alignments allows the initial TryAllAlign with the try2 costfcn (which includes the 4 disulfide bridges) to find an alginment to 1hykA that it likes. It twiddles it a bit afterwards, but doesn't change it very much. (see decoys/miniAGRP.try2-opt2.pdb.gz) The miniAGRP.t2k.w0.5.saves or miniAGRP.t04.w0.5.saves files could be used to suggest conservative substitutions for each position, perhaps giving Rosetta just the first 5-10 residues in each position (omitting AVLISTGFYK VILAFTMSYK RPKESADTLG LHVAIKRTES HQEKRSDALN EQKDARSTNG SATGEDKNRP AVLISTGFYK LIVFAMTYKS GASDNEKTRP HQEKRSADLN QHEKRSDALN LVIAFTMSYK PASEKGDLTV AVLISTGFYK AVLISTGFYK DNESGKATRQ PASEKGDLTV AVLSITGEKF ADESKGNTRQ TSAVLEKDGN AVLISTGFYK YFLVAISTWK AVLISTGFYK RKEASQLTDG FLYIVATWSM FLYIVATWSM KNEDSRGATQ AVGSLTEKID FLYIVATSMW AVLISTGFKE YFLVAISTKE AVLISTGFKE RKEASLQDTG Wed May 25 17:08:37 PDT 2005 Kevin Karplus Based on both the results here and in ../asip/asip.t04-w0.5.saves I'd recommend trying a design with something like VALISTFP VILASTFM RPAEKSDNTQ LTVIAHFSY HQERKSDA EDNQKSARTP STANDE VALISTFP LKREAVNTSIQD GSTPADE PHSQEDKA QASEKLDNTRYH VILAPTMF PAS VALISTFP VALISTFP DNESK PQSAEKD AST AVDESLKTRQ TSLFVAIM VALISTFP YFLQHEASKV # maybe force Y VALISTFP R F F NKRDSEATQP ATSEQKDR FLVAITSNY VALISTFP YSLTRNAEKV # maybe force Y VALISTFP RKASPETDN If other residues are believed to be part of the active site, they may need to be forced also. The design should start with the native residues, except for replacing the CYS with VAL, then optimize rotamers before beginning the design. Sat Jul 16 14:35:18 PDT 2005 Kevin Karplus Further work will be in the subdirectory design/ which attempts to use neural nets to get plausible sequences. See design/README for notes.