My apologies, a correction to my last posting:
On my final point concerning the calculation of D and the content of the
FC column in the MTZ file output by Refmac: I actually used the FC_ALL
column for all calculations, not the FC column as I implied. I use the
FC_ALL, PHIC_ALL columns for SF calculations because I believe that
these contain Fcalc including the bulk solvent background, whereas I
have been assuming that the FC, PHIC columns contain Fcalc omitting this
correction. An mtzdump of the file for the (0,13,10) reflection shows:
H K L FOM F SIGF FreeR_flag FC PHIC
0 13 10 0.89 383.48 7.05 14.00 98.96 99.76
FC_ALL PHIC_ALL FWT PHWT DELFWT PHDELWT
182.87 129.48 501.58 129.48 159.36 129.48
From the phase values PHIC, PHIC_ALL, PHWT & PHDELWT above it is clear
that FWT and DELFWT have been computed using FC_ALL not FC, so that's
why I actually used the FC_ALL value (182.9) in my example.
Cheers
-- Ian
> -----Original Message-----
> From: [log in to unmask]
> [mailto:[log in to unmask]] On Behalf Of Ian Tickle
> Sent: 30 December 2008 23:24
> To: [log in to unmask]
> Subject: Published derivation of mFo-DFc formula?
>
> All, something for you to think about over the Study Weekend - sorry I
> can't be there with you this time :-( :
>
> I was recently asked whether a formal derivation of the expression we
> use for the 'minimally biased' difference Fourier coefficient, i.e.
> delta-Fm = mFo-DFc, is published anywhere, and I was forced to admit
> that I didn't know of any such publication. It's certainly
> mentioned in
> program documentation (SIGMAA, REFMAC, etc), and also in online course
> material, but I could find no proof of the formula. Then I
> got thinking
> about how one would go about deriving it, and that opened up
> a whole new
> can of worms!
>
> First, if we define the coefficient Fm for the 'minimally biased'
> Fourier (as derived by Read, AC 1986, A42, 140) then:
>
> Fm = 2mFo - DFc for acentrics,
> = mFo for centrics.
>
> I'm reasonably happy with Randy's proof of this. My first question is
> why isn't the difference map coefficient given by:
>
> delta-Fm = Fm - DFc
> = 2(mFo - DFc) for acentrics,
> = mFo - DFc for centrics.
>
> This would reflect the fact that provided the amount of missing
> structure is small, peaks in a centrosymmetric difference Fourier show
> up around the full expected height, whereas those in a
> non-centrosymmetric one show up at about half-height, so you need to
> double the Fo-Fc difference. Indeed this is the whole
> rationale for the
> formulae Fo = Fc + (Fo-Fc) for centrics and 2Fo-Fc = Fc + 2(Fo-Fc) for
> acentrics.
>
> AFAIK all programs use the same expression delta-Fm = mFo-DFc for both
> acentrics and centrics, and looking at the SIGMAA code it
> certainly does
> (code snippet edited for clarity):
>
> IF (IC.EQ.1) THEN
> C
> C CENTRIC DATA: EITHER M*FO OR M*FO-D*FC
> C
> FOUT = W*FO
> FOUTD = FOUT -DLUZ*FC
> ELSE
> C
> C NON-CENTRIC DATA: EITHER 2*M*FO-D*FC OR M*FO-D*FC
> C
> FOUT = 2.0*W*FO - DLUZ*FC
> FOUTD = W*FO - DLUZ*FC
> ENDIF
>
> Incidentally there's the following code snippet in the centric part of
> the IF..ELSE..ENDIF block above:
>
> cv C. Vonrhein Jul 5 1999
> c
> c if we don't have FP set it to -D*Fc
> c
> IF (LOGMSS(IFO)) FOUT = -DLUZ*FC
>
> Shouldn't this be FOUT = DLUZ*FC (so that FOUTD = 0), if not then why
> the minus sign? And shouldn't the same test be applied to acentrics?
>
> Then I got thinking even more (always dangerous!), and so my second
> question is why isn't the difference map coefficient given instead by:
>
> delta-Fm = Fm - Fc
> = 2mFo - (1+D)Fc for acentrics,
> = mFo - Fc for centrics.
>
> My rationale for this is that the difference map is supposed
> to tell you
> what changes you need to make to the current model (represented by Fc)
> in order to obtain the minimally biased model (represented by
> Fm). The
> point is that DFc does not represent the current model: as I
> understand
> it, it's a partial structure factor representing the part of
> the current
> model that's correct, the remaining part being random error.
>
> I modified SIGMAA and looked at some difference maps computed
> using the
> new expressions, and I have to admit they're not strikingly different,
> though there are some differences from the maps computed using the
> standard version. That could be because my D's are close to
> 1, in which
> case there would be no difference: maybe there would be a
> bigger effect
> if my D's were significantly less than 1 (D of course can't
> be < 0 or >
> 1). Even so it's always nice to use expressions that can be justified
> theoretically. A formal derivation of the mFo-DFc expression would
> settle this, and I would be happy to accept whatever result such a
> derivation gives, assuming that one exists.
>
> Then just for fun I looked at the FWT & DELFWT columns in an MTZ file
> output by Refmac, and this is where the worms really get out
> of the can!
> I picked an acentric and a centric reflection at random (SG P21):
>
> hkl = (0,13,10) FP=383.5 FC=182.9 FWT= 501.6 DELFWT= 159.4
> FOM=0.89
> hkl = (-5,0,11) FP= 82.4 FC=410.9 FWT=-271.6 DELFWT=-341.2
> FOM=0.85
>
> Note that the -ve values of FWT & DELFWT arise because the
> phases differ
> by 180 from phi(calc).
>
> First I checked for consistency of the FOM values since for acentric:
>
> m = ((2mFo-DFc)-(mFo-DFc))/Fo = mFo/Fo = (501.6-159.4)/383.5 = 0.89 so
> OK.
>
> For centric case m = mFo/Fo = -271.6/82.4 = -3.3 so not OK!
>
> However if I use the acentric formula for the centric case: m =
> (-271.6-(-341.2))/82.4 = 0.845 . So Refmac is either not detecting
> centrics correctly or is using the wrong formula.
>
> Then I checked for consistency of the D values since for acentrics:
>
> D = ((2mFo-DFc)-2(mFo-DFc))/Fc = DFc/Fc
>
> For the acentric case D = (501.6-2x159.4)/182.9 = 1.000 .
>
> I've already established it's using the acentric formulae for
> centrics,
> so for the centric case:
>
> D = (-271.6-2(-341.2))/410.9 = 1.000 .
>
> In fact D computes to exactly 1 for every reflection, so something is
> not right; either D isn't being computed correctly, or a possible
> explanation is that the FC column contains DFc instead of Fc. This
> would be unfortunate since we use Refmac to compute structure factors
> containing the solvent background contribution (which SFALL doesn't of
> course).
Disclaimer
This communication is confidential and may contain privileged information intended solely for the named addressee(s). It may not be used or disclosed except for the purpose for which it has been sent. If you are not the intended recipient you must not review, use, disclose, copy, distribute or take any action in reliance upon it. If you have received this communication in error, please notify Astex Therapeutics Ltd by emailing [log in to unmask] and destroy all copies of the message and any attached documents.
Astex Therapeutics Ltd monitors, controls and protects all its messaging traffic in compliance with its corporate email policy. The Company accepts no liability or responsibility for any onward transmission or use of emails and attachments having left the Astex Therapeutics domain. Unless expressly stated, opinions in this message are those of the individual sender and not of Astex Therapeutics Ltd. The recipient should check this email and any attachments for the presence of computer viruses. Astex Therapeutics Ltd accepts no liability for damage caused by any virus transmitted by this email. E-mail is susceptible to data corruption, interception, unauthorized amendment, and tampering, Astex Therapeutics Ltd only send and receive e-mails on the basis that the Company is not liable for any such alteration or any consequences thereof.
Astex Therapeutics Ltd., Registered in England at 436 Cambridge Science Park, Cambridge CB4 0QA under number 3751674
|