protein backbone assignment relying on known structure and low quality spectra

[Evgeny Fadeev]

Hi,

has anyone heard of work where protein resonances were assigned using
a homology structural model or known structure?

basically when NMR data is incomplete/low quality.
and "brute force" assignment would fail.

Maybe you could suggest somebody who might give insight or have an
idea.

Thanks.

Evgeny.

[Scott Robson]

I don't know of anyone doing this, but I would think the way to
proceed would be with NOESY data.

If the data is poor or incomplete because of relaxation (not exchange)
then the NOESY data may look better. Map the assignments you have made
onto the NOESY spectra, and then use what you do know about secondary
structure from the homology model to make guesses at assignments for
connecting residues you can't assign in scalar coupled experiments.

This assumes you have enough assignments in the first place to do some
sequence specific assignments.

S/

--
Scott A. Robson
Postdoctoral Research Fellow
UCLA Department of Chemistry & Biochemistry

[Evgeny Fadeev]

Thanks. noesy in combination with shift predictions and structure
sounds like a possibility.

one needs to put it all into one software package otherwise it would be tough
go manually between lists of approximate shifts, structure display and spectra.
or maybe it might even work without human intervention.

Rob just metioned this:
>The Wishart group has developed programs that may be useful.
>http://redpoll.pharmacy.ualberta.ca/projects/nmr.html
Copy from the website: "To date we have succeeded in developing a
computer program (called PHIPSI2CS) that accurately and rapidly
calculates backbone and sidechain 15N, 1H and 13C chemical shifts from
3D coordinate data. Tests indicate tha this program is the most
accurate chemical shift prediction method so far developed. We have
also developed a program (called PROTEIN BUILDER) that builds 3D
structures from a combination of J-coupling and chemical shift data.
This program was used to construct the 3D structures (independent of
NOE information) of five cyclic gramicidin S peptides (published in
the April issue of Nature Structural Biology). This is the first
successful demonstration of NMR-based 3D structure generation without
the use of explicit NOE information."