Hi, everyone,
Wladek has informed me that the profile fitting radius should be set by
default to 1/8 the size of the detector, which would correspond to 40 mm
on our MAR 325 CCD. They will change that in their next release. And
thanks for the incredibly quick response from everyone, it has not even
been a day, and not even a working day, on a holiday weekend!
Engin
meindert lamers wrote:
> Hi Engin,
>
> I have seen the same with HKL2000, especially with weak diffracting
> crystals.
> Don't know if this is correct, but another way to overcome this is to
> lower the value of "error density" during integration.
>
> I've seen the value for error density vary between different sites
> (beamlines). In one particular case the value was 3.3, which resulted
> in a completeness of 1.1% (!) in the outer shell. Mosflm/Scala
> reported a completeness of ~70%
>
> Trying different values for error density gave me:
> value compl overall / outer shell
> 3.3 61.4 / 1.1
> 2.0 67 / 3.7
> 1.0 76 / 12
> 0.3 88 / 47
> 0.1 94 / 83
>
> I wouldn't know how to choose the correct value for error density
> (probably different for each detector), so I looked at the quality of
> the electron density maps to decide which value to use.
>
> Meindert
>
> ************************************
> Meindert H. Lamers, PhD
> Kuriyan lab - Molecular and Cellular Biology
> University of California, Berkeley
> 527 Stanley Hall, QB3
> Berkeley, CA 94720-3220
> Tel: (510) 643 0164 (Fax: 2352)
> ************************************
>
> On Apr 11, 2009, at 6:11 PM, Engin Ozkan wrote:
>
>> Hi,
>>
>> The problem was the profile test failing as suggested by Wladek and
>> Lothar (thanks to both, for continuous teaching since grad school). I
>> was using a profile fitting radius of 25 (not the default 10), so
>> that's why I assumed profile fitting was not the problem. But now, I
>> have played with the profile fitting radius (PFR), and below are the
>> numbers.
>>
>> PFR Compl% Compl% (high res bin)
>> 25 81 63
>> 50 96 80
>> 60 98 86
>>
>> And with hindsight, it was very embarrassing to find in my log file
>> "count of observations deleted due to zero sigma or profile test",
>> which tells me the problem. And again, because mosflm was giving me
>> 100% complete datasets, it had to be a scalepack-specific
>> implementation.
>>
>> Interestingly, at a PFR of 50, a single profile covers almost a
>> fourth of the image. Any input on that one?
>>
>> Engin
>>
>> ----- Original Message -----
>> From: "Wladek Minor" <[log in to unmask]>
>> To: [log in to unmask]
>> Sent: Saturday, April 11, 2009 5:11:40 PM GMT -08:00 US/Canada Pacific
>> Subject: Re: [ccp4bb] HKL2000 and incompleteness
>>
>> Try to increase profile fitting radius.
>>
>> WM
>>
>>
>> ------Original Message------
>> From: Ted Erickson
>> Sender: CCP4 bulletin board
>> To: [log in to unmask]
>> ReplyTo: Ted Erickson
>> Subject: Re: [ccp4bb] HKL2000 and incompleteness
>> Sent: Apr 11, 2009 7:56 PM
>>
>>> So before I lose it because of HKL2000, I have two questions, and I'd
>> appreciate any answers to either:
>>> 1. Can there be another reason for incompleteness that I am missing
>>> (other
>> than blind region at higher res., overlaps, overloads, and "not
>> enough many
>> frames")?
>>
>> Another explanation for incompleteness could be due to crystal
>> packing that
>> results in an asymmetric diffraction pattern. The discrepancy between
>> HKL2000 and mosflm maybe a result of the resolution cutoff, however,
>> this is
>> only a speculation.
>>
>> Ted
>> www.P212121.com
>>
>>
>>
>> Sent via BlackBerry by AT&T
>
|