[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: HL coefficients from DM
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html ***
Dear Ana,
in my experience, the HL coefficients after any kind of density modification
are vastly overestimated (to my knowledge, there is yet no program with a
good statistical treatment of density modification). If you use
maximum-likelihood refinement with these HL coefficients, it will be
dominated by them and nothing really happens during refinement. In REFMAC
you can downweight the HL coefficients, but at least in my hands, that
never really worked satisfactorily. However, I've made VERY good experiences
using directly the HL coefficients of your heavy atom refinement and phasing
program (especially from SHARP!) in maximum-likelhood refinement.
Reasonably well weighted observed phases will help you a lot in completing
your initial partial model!
Good luck,
Dirk.
*******************************************************
* Dirk Kostrewa phone: +41-61-688-7750 *
* F.Hoffmann-La Roche Ltd. fax: +41-61-688-7408 *
* PRPC-S B65/R312 *
* CH-4070 Basel, Switzerland *
* e-mail: dirk.kostrewa@roche.com *
*******************************************************
> -----Original Message-----
> From: Ana Gonzalez [SMTP:Ana@embl-hamburg.de]
> Sent: Wednesday, July 14, 1999 13:29
> To: ccp4bb@dl.ac.uk
> Subject: HL coefficients from DM
>
> *** For details on how to be removed from this list visit the ***
> *** CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html ***
>
> Hello,
>
> Does anyone ever use the HL coefficients from DM? If so, do they look
> reasonable?
>
> The story behind the question:
>
> I have a very nice set of MAD phases, which I decided to use in the
> structure refinement. I used refmac with the phase and figure of merit
> after solvent flattening and it worked fairly well. Then I felt curious
> (yes, I know what they say about curiosity) about whether using the HL
> coefficients would give even better results.
> Unfortunately refmac did not like the HLDMs that DM generated. Looking at
> the mtz file I get this:
>
> Min Max Mean Abs Mean
> -2056.0 1825.0 94.20 53.45 HADM
> -1572.7 1491.2 94.90 53.45 HBDM
> -2146.6 2705.8 80.42 53.45 HCDM
> -2424.1 3345.0 80.99 53.45 HDDM
>
> I found this surprising, since I would expect the HLs to be quite small -
> in fact, the ones generated by the phasing program look like this (same
> columns as above):
>
> -57.0 30.9 0.00 0.46 HA
> -45.2 44.2 0.00 0.46 HB
> -14.4 9.1 0.00 0.04 HC
> -8.8 7.5 0.00 0.04 HD
>
> Could this be a bug on DM? (I have also seen this dor another case, SIRAS
> phases, so it does not look like it is a pathology associated to my data)
>
> Cheers,
>
> Ana