[sharp-discuss] Rcullis(anomalous) for SAD data?
Marc SCHILTZ
schiltz@globalphasing.com
Tue, 11 Mar 2003 16:12:16 +0000
Dear Ed,
In SHARP, anomalous lack-of-closure variance (LOCV) for a
reflection is defined as:
LOCV_ano(hkl) = <(DANO_obs - DANO_calc)**2>
where:
DANO_obs = |Fph+|_obs - |Fph-|_obs
DANO_calc = Scale*(|Fp* + Fh+_calc| - |Fp* + Fh-_calc|)
Fp* = 'trial' point for protein structure-factor
Fh+_calc / Fh-_calc = structure factors computed from
heavy-atom model
(in the case of SAD, 'heavy atom' means anomalous scatterer)
The average < ... > in the above formula is a weighted average
over all possible Fp* on the complex plane. Remember Fp* is
unknown and the philosophy of SHARP is to integrate over all
possible values of Fp* (a process termed 'marginalization').
More intuitively, LOCV_ano represents the (squared) part of
observed anomalous differences that can not be explained by
the heavy atom model (i.e. which may be termed non-isomorphism).
Now, anomalous R-Cullis is defined as:
SQRT[Sum_{hkl} LOCV_ano(hkl)]
RCullis_ano = --------------------------------
SQRT[Sum_{hkl} DANO_obs(hkl)**2]
And anomalous Phasing Power is:
SQRT[Sum_{hkl} <DANO_calc(hkl)**2>]
PhasPow_ano = -----------------------------------
SQRT[Sum_{hkl} LOCV_ano(hkl)]
where again, the average < ... > is a weighted average
over all possible Fp* on the complex plane.
EXCEPT for the integration over Fp*, these relations are
equivalent to:
rms(DANO_calc - DANO_obs)
RCullis_ano = ---------------------------
rms(DANO_obs)
rms(DANO_calc)
PhasPow_ano = ---------------------------
rms(DANO_calc - DANO_obs)
--
================================================================
* Marc SCHILTZ *
* Global Phasing Ltd. *
* Sheraton House *
* Castle Park *
* Cambridge CB3 0AX *
* UK *
* email:schiltz@GlobalPhasing.com *
* Tel: +44-(0)1223-353033 Fax: +44-(0)1223-366889 *
================================================================