Two suggestions:
What happens if you tell HKL2000 that you want more bins in the scaling so you get a different number of reflections per bin? My personal preference for data shown like your's is to have more than 20 (!) bins because there are an enormous number of reflections available to do a decent job on the scaling. This may help because the statistics inside each bin will be different.
As far as I can think about this, it does not matter if you multiplied all intensities AND sigma values by an arbitrary number, so you avoid rounding down of an absolute number. You have to do this carefully so the I/sigma values don't change. Maybe this is cheating and someone will object. After the data have been scaled the result should be the same, minus the omissions you see?
Mark
-----Original Message-----
From: Nat Echols <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Sent: Fri, 10 Feb 2006 19:22:35 -0800 (PST)
Subject: [ccp4bb]: Re: reflection file conversion
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
I believe I figured out the reason (but not the solution) to this problem:
> 2) This is unrelated to SHELX: I integrated and scaled in HKL2000 and then > converted to MTZ format. All of the CCP4 programs say that my data is > incomplete, but the scalepack log file says I'm 100% complete in all shells. > I looked at the MTZ file with mtzdump and it says I'm missing data around > 1.7-1.5A, but not in higher-resolution shells. Anyone have an idea what's > going on?
The problem is that when I merge the high-res and low-res wedges, HKL2000 scales the intensities such that the range starts very low:
Shell Lower Upper Average A! verage Norm. Linear Square
limit Angstrom I error stat. Chi**2 R-fac R-fac
50.00 2.11 38.3 0.5 0.3 1.540 0.041 0.049
2.11 1.68 9.5 0.2 0.1 1.112 0.077 0.081
1.68 1.46 3.7 0.1 0.1 1.063 0.097 0.132
1.46 1.33 2.2 0.1 0.1 0.997 0.084 0.087
1.33 1.23 1.6 0.1 0.1 0.963 0.109 0.114
1.23 1.16 1.3 0.1 0.1 0.938 0.134 0.126
1.16 1.10 1.0 0.1 0.1 0.865 0.165 0.155
1.10 1.06 0.6 0.1 0.1 0.754 0.253 0.239
1.06 1.02 0.4 0.1 0.1 0.674 0.374 0.363
1.02 0.98 0.2 0.1 0.1 0.627 0.543 0.484
This is not a problem when processing the high-res data separately; I have, on the other hand, observed this result for other proteins where I had high- and low-res wedges.
The missing reflections arise from HKL2000 setting sigma(I) to 0.0 in cases where it was very low - I assume this is just rou! nding down. This has not happened with similar data in the past, but f or this crystal (my best for that protein!) more than 10% of the reflections have this problem. scalepack2mtz treats all reflections with sigI=0 as missing (based on the output, and the source code - this is ccp4 5.0.2).
I guess I have three choices:
1) Tell HKL2000 to *not* scale everything down so much. (Is this even possible? FYI, ighest intensity is 1665, which is the only intensity above 1000.)
2) Modify scalepack2mtz to keep reflections with sigI=0. I have no idea what the effect would be on other programs of zero sigma values.
3) Change sigI to 0.1 in all cases where it's 0.0. I suspect this is probably a bad idea.
Any thoughts?
thanks,
Nat
*** CCP4 home page http://www.ccp4.ac.uk ***
I believe I figured out the reason (but not the solution) to this problem:
> 2) This is unrelated to SHELX: I integrated and scaled in HKL2000 and then > converted to MTZ format. All of the CCP4 programs say that my data is > incomplete, but the scalepack log file says I'm 100% complete in all shells. > I looked at the MTZ file with mtzdump and it says I'm missing data around > 1.7-1.5A, but not in higher-resolution shells. Anyone have an idea what's > going on?
The problem is that when I merge the high-res and low-res wedges, HKL2000 scales the intensities such that the range starts very low:
Shell Lower Upper Average A! verage Norm. Linear Square
limit Angstrom I error stat. Chi**2 R-fac R-fac
50.00 2.11 38.3 0.5 0.3 1.540 0.041 0.049
2.11 1.68 9.5 0.2 0.1 1.112 0.077 0.081
1.68 1.46 3.7 0.1 0.1 1.063 0.097 0.132
1.46 1.33 2.2 0.1 0.1 0.997 0.084 0.087
1.33 1.23 1.6 0.1 0.1 0.963 0.109 0.114
1.23 1.16 1.3 0.1 0.1 0.938 0.134 0.126
1.16 1.10 1.0 0.1 0.1 0.865 0.165 0.155
1.10 1.06 0.6 0.1 0.1 0.754 0.253 0.239
1.06 1.02 0.4 0.1 0.1 0.674 0.374 0.363
1.02 0.98 0.2 0.1 0.1 0.627 0.543 0.484
This is not a problem when processing the high-res data separately; I have, on the other hand, observed this result for other proteins where I had high- and low-res wedges.
The missing reflections arise from HKL2000 setting sigma(I) to 0.0 in cases where it was very low - I assume this is just rou! nding down. This has not happened with similar data in the past, but f or this crystal (my best for that protein!) more than 10% of the reflections have this problem. scalepack2mtz treats all reflections with sigI=0 as missing (based on the output, and the source code - this is ccp4 5.0.2).
I guess I have three choices:
1) Tell HKL2000 to *not* scale everything down so much. (Is this even possible? FYI, ighest intensity is 1665, which is the only intensity above 1000.)
2) Modify scalepack2mtz to keep reflections with sigI=0. I have no idea what the effect would be on other programs of zero sigma values.
3) Change sigI to 0.1 in all cases where it's 0.0. I suspect this is probably a bad idea.
Any thoughts?
thanks,
Nat
Try the New Netscape Mail Today!
Virtually Spam-Free | More Storage | Import Your Contact List
http://mail.netscape.com
