How good is good (was Re: Comment to S.Blocklehurst (Re: Well, ...))

Arne Elofsson arne at
Fri May 5 13:31:33 EST 1995

In article <3odceg$dah at> kpmurphy at (murphy kenneth p) writes:

> This brings up a question I've always had regarding structure prediction.  How
> close do you have to be to the crystal structure to say you predicted it?  When
> you say 1.0 A, do you mean for all atoms or just backbone?  What's the rmsd for
> NMR and xtal structures of the same protein if you include all atoms and not
> just backbone?

I would be happy with anything that could help me understand the biological 
function of the protein, even if it is 10 A away.

For the first RMSD is not a very good measure but I do not know anything better.

Let me think loudly (or recall from memory):

1) Differences between different BPTI forms (3-9pti excluding 8pti) is less than
   1 A for backbone and prrobably between 1 and 1.5 A
2) Root mean square fluctuations (tempfactors) are less than 1 A.
3) Best MD simulations are close to 1A rmsd for all atoms.
4) IL-4 (?) differences between NMR and X-ray are close to 1 A (slightly more a guess)

So this would mean that I would be satisfied to get  1-2 A accuracy.

But I think we are a long way from there.


               From: Arne Elofsson
         Email: arne at   			

More information about the Proteins mailing list