Re: [ccp4bb]: unusually high solvent content

[Paul Emsley <paule@chem.gla.ac.uk>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***


Although somewhat tangentially related to the original question I
though that it may be of interest to some, that the last time I
checked the solvent content of protein structures published in the PDB
(which was a while ago now I must admit (this was an initial attempt
to automate the correct estimation of the NCS (a task which I continue
today)) the average solvent content of protein crystals was 47.5% with
a standard deviation of 9.6%, but what was particularly surprising to
me at least was that the distribution of solvent contents was
gaussian-like and if one can use that to predict probabilities of high
solvent contents (and one should of course take such calculations with
a pinch of metaphorical salt since often real-world data often show
considerably higher probability densities at higher Z values than one
would have suspected from a purely gaussian distribution) the value
you quote of the "limiting" mathews coefficient (3.5) corresponds to a
solvent content of 65% (at least according to my calculations) and a
p-value of 3.5% (for proteins having solvent content greater than
that) which means that if one were to calculate the range between
which the matthews coefficient were just as likely to lie as not
(with no other information) then that range would be 2.03 to 2.67. 
I hope that answers to some extent one of your questions.

It is left as an amusing(?) holiday exercise for the vodka-embibing
reader to appy the Kolmogorov-Smirnov test to protein solvent
contents.

HTH,

Paul.
-- 
Paul.Emsley@chem.gla.ac.uk   
 http://www.chem.gla.ac.uk/~paule