2009-12-17T21:57:16Z

Yphuang:

AutoStructure [1,2,3] is an automated NOESY assignment engine, which uses a distinct bottom-up topology-constrained approach for iterative NOE interpretation and structure determination. AutoStructure first builds an initial chain fold based on
intraresidue and sequential NOESY data, together with characteristic NOE
patterns of secondary structures, including helical medium-range NOE
interactions and interstrand b-sheet NOE interactions, and unambiguous long-range
NOE interactions, based on chemical shift matching and NOESY spectral symmetry
considerations. NOESY cross peaks that cannot be uniquely assigned using these
methods are not used in the initial structure calculations.  Once initial structures are generated and validated, additional NOESY cross peaks are iteratively assigned using the intermediate 3D structures and contact maps, together with knowledge of high-order topology constraints of alpha-helix and beta-sheet packing geometries [4]. This protocol, in
principle, resembles the method that an expert would utilize in manually
solving a protein structure by NMR.   

This ‘bottom up’ strategy is quite different from the “top down”
strategies used by the alternative programs CANDID and ARIA, which rely on
“ambiguous constraints”.  For NOESY spectra with poor signal-to-noise ratios, such automatically assigned ‘ambiguous constraint” sets may not include any true NOESY assignments, and result in distortions of the protein structure which can be avoided by the “bottom up” approach of AutoStructure [4].  CANDID/CYANA also uses a ‘network anchoring” approach similar to, but less comprehensive than, the topology-constrained approach used by AutoStructure [4].  For these reasons, some users may prefer to use both AutoStructure and CANDID/CYANA or ARIA in parallel to assess potential errors in automated NOESY cross peak assignments [83]. Fig. 4 shows AutoStructure results for three different human protein NMR
test data sets: FGF-2, IL-13 and MMP-1, ranging in size from 113 to 169
amino-acid residues. The mean coordinate differences between structures
determined by AutoStructure and by manual analysis (0.5 to 0.8 Å for backbone
atoms of ordered residues) demonstrate good accuracy of these automated
methods.  The input data for AutoStructure are: (i) resonance assignment table, (ii) 2D, 3D, and/or 4D NOESY peak lists, (iii) list of scalar coupling, RDC and slow amide exchange data. AutoStructure generates distance constraint lists and utilizes the programs DYANA/CYANA, Xplor for 3D structure
generation on a Linux-based computer cluster.

[[Image:AS.jpg|left|307x229px|Ribbon diagrams of representative structures of FGF-2, MMP-1 and IL-13 proteins used for the validation of the AutoStructure process]]

Fig. 1. Ribbon diagrams of representative structures of FGF-2, MMP-1 and IL-13 proteins used for the validation of the AutoStructure process: (a) final structures from AutoStructure using XPLOR for stucture generation, (b) manual-analyzed structures deposited in PDB, analyzed using the same NMR data set, (c) structures determined by X-ray crystallography or third NMR group. Tabulated on the right are mean coordinate differences (Å) in secondary structure regions for backbone atoms between structures (a), (b) and (c).

2009-12-17T21:27:21Z

Yphuang: Created page with ''''AutoStructure – An automated NOESY assignment engine.''' '''<span style="font-size:11…'

'''AutoStructure – An automated NOESY assignment engine.'''

''''''AutoStructure [1,2,3], written in C/C++ and Perl/Tk, uses a distinct bottom-up''' topology-constrained approach for iterative NOE interpretation and structure determination. AutoStructure first builds an initial chain fold based on
intraresidue and sequential NOESY data, together with characteristic NOE
patterns of secondary structures, including helical medium-range NOE
interactions and interstrand b-sheet NOE interactions, and unambiguous long-range
NOE interactions, based on chemical shift matching and NOESY spectral symmetry
considerations. NOESY cross peaks that cannot be uniquely assigned using these
methods are not used in the initial structure calculations.  Once initial structures are generated and validated, additional NOESY cross peaks are iteratively assigned using the intermediate 3D structures and contact maps, together with knowledge of high-order topology constraints of alpha-helix and beta-sheet packing geometries [82]. This protocol, in
principle, resembles the method that an expert would utilize in manually
solving a protein structure by NMR.  

<o:p> </o:p>

[[Image:|Text Box: Fig. 4. Ribbon diagrams of representative structures of FGF-2, MMP-1 and IL-13 proteins used for the validation of the AutoStructure process: (a) final structures from AutoStructure using XPLOR for stucture generation, (b) manual-analyzed structures deposited in PDB, analyzed using the same NMR data set, (c) structures determined by X-ray crystallography or third NMR group. Tabulated on the right are mean coordinate differences (Å) in secondary structure regions for backbone atoms between structures (a), (b) and (c). ]][[Image:]]

















This ‘bottom up’ strategy is quite different from the “top down”
strategies used by the alternative programs CANDID and ARIA, which rely on
“ambiguous constraints”.  For NOESY spectra with poor signal-to-noise ratios, such automatically assigned ‘ambiguous constraint” sets may not include any true NOESY assignments, and result in distortions of the protein structure which can be avoided by the “bottom up” approach of AutoStructure [4].  CANDID/CYANA also uses a ‘network anchoring” approach similar to, but less comprehensive than, the topology-constrained approach used by AutoStructure [4].  For these reasons, some users may prefer to use both AutoStructure and CANDID/CYANA or ARIA in parallel to assess potential errors in automated NOESY cross peak assignments [83]. The AutoStructure program
has been used to determine 3D structures for most of the > 300 proteins
solved by the NESG consortium.<o:p></o:p>

<o:p> </o:p>

Fig. 4 shows AutoStructure results for three different human protein NMR
test data sets: FGF-2, IL-13 and MMP-1, ranging in size from 113 to 169
amino-acid residues. The mean coordinate differences between structures
determined by AutoStructure and by manual analysis (0.5 to 0.8 Å for backbone
atoms of ordered residues) demonstrate good accuracy of these automated
methods.  The input data for AutoStructure are: (i) resonance assignment table, (ii) 2D, 3D, and/or 4D NOESY peak lists, (iii) list of scalar coupling, RDC and slow amide exchange data. AutoStructure generates distance constraint lists and utilizes the programs DYANA/CYANA, Xplor for 3D structure
generation on a Linux-based computer cluster.