I am not mistaken what went wrong in my case is that
in resolve.mtz there are the columns FP and SIGFP which are not the experimental F anymore
but they are modified ones (solvent flattened, ncs etc...).
Clearly refmac5 uses them as Fo for refinement, refining Fc against Fc (an extreme case of over-refinement?),
or in this case better say tautology.
stefano
----------------------------------------
> Date: Mon, 31 Mar 2008 16:50:13 +0100
> From: [log in to unmask]
> Subject: Re: [ccp4bb] too good R/Rfree with resolve
> To: [log in to unmask]
>
> On Sun, Mar 30, 2008 at 07:08:14PM +0100, Partha Chakrabarti wrote:
>> Is it by any chance that the FOMs were highly overestimated and that
>> creates a problem with Maximum likelihood? That sort of reminds me of
>> what I had heard for SHARP-solomon in a couple of instances..
>
> The FOMs are never used in anything downstream from experimental
> phasing and/or density modification - unless you calculate maps, but
> even then you usually use some map coefficients directly. And to
> describe the phase probability you should ideally use
> Henrickson-Lattmann coefficients I guess.
>
> Anyway, if you want to use experimental phase information in
> refinement (good idea), make sure to use the Henrickson-Lattmann
> coefficients from the phasing program. The HLs after density
> modification are usually over-optimisitic (at least for classical
> density modification programs like SOLOMON or DM).
>
> I must admit I've never heard suspicions about the SHARP-solomon FOMs
> before ... I could imagine people being generally supicious about FOMs
> from any density modification program? I have very good experiences
> using the HLs from SHARP afterwards during refinement. If you have
> more details about this, I'd be very interested to hear!
>
> Cheers
>
> Clemens
>
>
>>
>> Any insight would be highly appreciated..
>> Regards, Partha
>>
>>
>> On Fri, Mar 28, 2008 at 2:35 PM, Garib Murshudov wrote:
>>> make sure that you are using the original observation (Fobs and
>>> corresponding sigmas) not that produced by density modification (e.g.
>>> solve resolve) programs.
>>>
>>> Garib
>>>
>>>
>>>
>>> On 28 Mar 2008, at 10:43, stefano ricagno wrote:
>>>
>>> > Dear CCP4bb readers,
>>> > this is my problem:
>>> > I solved a structure by MR: the solution was easily found (molrep,
>>> > phaser and balbes found always the same one), density looked
>>> > generally reasonable (however in several places it was dubious) but
>>> > R/Rfree were stuck at 42/47%.
>>> > Then I tried some density modifications, resolve worked
>>> > spectacularly the density became wonderful and several parts which
>>> > were not in the model appeared.
>>> > So I finished to build the model and everything looked good.
>>> > The problem is now for a structure at 2.8 Å resolution I have R/
>>> > Rfree of 8/9.5% respectively, which is clearly too good. Checking
>>> > refmac log file it looked to me that refmac uses all the
>>> > reflections in the .mtz file (that is as many reflections as before
>>> > the resolve run).
>>> > Ideas?
>>> >
>>> > thanks
>>> >
>>> > Stefano
>>> >
>>> > ps this is my last refmac:
>>> >
>>> > 2 mol in the AU, no NCS used, weighting term 0.01, no tls.
>>> >
>>> > Ncyc Rfact Rfree FOM -LL -LLfree rmsBOND
>>> > zBOND rmsANGL zANGL rmsCHIRAL $$
>>> > $$
>>> > 0 0.0943 0.1051 0.937 102149. 5673.2 0.0070
>>> > 0.298 1.202 0.546 0.077
>>> > 1 0.0910 0.1024 0.940 101442. 5642.4 0.0058
>>> > 0.252 1.091 0.481 0.073
>>> > 2 0.0898 0.1019 0.940 101221. 5634.0 0.0055
>>> > 0.238 1.061 0.461 0.072
>>> > 3 0.0891 0.1016 0.941 101080. 5628.8 0.0052
>>> > 0.226 1.045 0.451 0.072
>>> > 4 0.0884 0.1011 0.942 100947. 5622.7 0.0050
>>> > 0.217 1.033 0.445 0.071
>>> > 5 0.0877 0.1006 0.942 100824. 5617.0 0.0049
>>> > 0.211 1.024 0.440 0.071
>>> > 6 0.0872 0.1001 0.943 100708. 5611.4 0.0047
>>> > 0.205 1.016 0.436 0.070
>>> > 7 0.0866 0.0997 0.943 100593. 5606.5 0.0047
>>> > 0.202 1.008 0.432 0.070
>>> > 8 0.0861 0.0992 0.944 100483. 5601.5 0.0046
>>> > 0.199 1.002 0.429 0.069
>>> > 9 0.0857 0.0989 0.944 100391. 5597.5 0.0045
>>> > 0.196 0.996 0.426 0.069
>>> > 10 0.0852 0.0985 0.944 100313. 5593.5 0.0045
>>> > 0.193 0.990 0.423 0.068
>>> >
>>> > _________________________________________________________________
>>> > Explore the seven wonders of the world
>>> > http://search.msn.com/results.aspx?q=7+wonders+world&mkt=en-
>>> > US&form=QBRE
>>> >
>>>
>>
>>
>>
>> --
>> MRC National Institute for Medical Research
>> Division of Molecular Structure
>> The Ridgeway, NW7 1AA, UK
>> Email: [log in to unmask]
>> Phone: + 44 208 816 2515
>>
>
> --
>
> ***************************************************************
> * Clemens Vonrhein, Ph.D. vonrhein AT GlobalPhasing DOT com
> *
> * Global Phasing Ltd.
> * Sheraton House, Castle Park
> * Cambridge CB3 0AX, UK
> *--------------------------------------------------------------
> * BUSTER Development Group (http://www.globalphasing.com)
> ***************************************************************
_________________________________________________________________
Explore the seven wonders of the world
http://search.msn.com/results.aspx?q=7+wonders+world&mkt=en-US&form=QBRE
|