Data handling

1
Title
Data handling Zbigniew DAUTER
Zbyszek (friends) Zbynius (MD)
Darling (MD sometimes) Zygmunt (GGD)
Speedcheck (KSW) Zibi (USA)
2
Name confusion
Confusing name (continuation) data
or data data was collected data were
collected (Polish is much easier)
3
Handle with care
k
h
Handle with care
Data
l
Fragile !
Do not abuse
4
H,k,l,I,sigma
h k l I s(I) very easy unfortunately
to collect these are also (KSW)
necessary
5
Uncomplete data
If data are not complete, there are problems
in - direct methods substructure
solution - Patterson methods substructure
solution Molecular Replacement - Fourier
maps additional features
6
Non-random missing refl.
Missing reflections are never random
unfinished rotation range - missing wedge
in orthorhombic symmetry 90o is not
always enough - only between two axial
orientations
unhappy Kevins duck
7
622 symmetry
In 622 symmetry if rotated around
6-fold 30o is enough if rotated around
2-fold 90o is necessary
8
Blind region
Blind region normally not
serious, except if - symmetry is low
(P1) - resolution is very high (atomic)
9
Overloads
Overloaded profiles detector dynamic
range, e.g. 216 limit (low resolution pass
may be necessary)
small number of missing strongest reflections
ruins Molecular Replacement (Gideon Davies, 93)
10
Accuracy criteria
Data accuracy criteria
Rmerge (Rsym, Rint )
__________________ I/s(I) -
generally gt 2, or 50 of reflections gt 3s
Redundancy the higher the better, but
beware of radiation damage
Shkl Si ltIgt - Ii Shkl Si Ii
11
Improved Rmerge
Rmerge - bad criterion from statistical
point of view (depends on redundancy)
Improved forms (sadly, rarely used) Rmeas
____________________ (Diederichs Karplus,
97) Rr.i.m. ______________________ (Weiss
Hilgenfeld, 97) Rp.i.m. __________________
(Weiss Hilgenfeld, 97)
Shkl n Si ltIgt - Ii Shkl (n-1) Si Ii
Shkl n1/2 Si ltIgt - Ii Shkl (n-1)1/2 Si Ii
Shkl Si ltIgt - Ii Shkl (n-1)1/2 Si Ii
12
Sigmas
I/s(I) is much better criterion 2-D
detectors do not measure individual X-ray
quanta but something proportional therefore
counting statistics is not valid ss must be
corrected for detector gain t-plot
_________ average 0, s.d. 1 c2
criterion - agreement with expectations
ltIgt - Ii s(I)
13
Invitation to York
Why was I invited to speak here ? Because
I learned from the masters ! I was a postdoc
in York 1981-1985
14
Woolfson
First in Physics with Michael
Woolfson
15
Guy Bakers
Then in Chemistry Learning from the
masters (Mirka too)
16
Zygmunt Casio
We had excellent computing facilities
Zygmunt refining haemoglobin
17
Long computations
Long computations Mirka helped too only
rarely the message would appear something
wrong consult the programmer (EJD)
18
Colin Eleanor
Cd- and Pb- Insulin data collected on a
Hilger-Watts diffractometer result - 5
punched paper tapes R3 became my favoured
space group - importance of reflection 472
SSM (Sort, Scale and Merge)
_
Colin Reynolds and EJD
19
Disappointments
Later disappointments Synchrotrons
everything too fast Autoindexing those
people (Rossmann, Kabsch, Otwinowski, Leslie)
spoiled the fun of fitting diffraction patterns
by hand Automatization now data
processing is like a black box Well, not
completely !
20
conclusion
Conclusion
21
Poor and sick
It is better to be rich and healthy
than poor and sick or to have complete and
accurate data rather than otherwise
gift from Thomas Schneider (has Italian
wife)
22
Cato the Elder
after Cato the Elder in the Roman
senate et memento delenda est
Carthago data collection involves a lot of
technical problems but is not a technicality
- last experimental step, later only
calculations it is good to engage the
brain
23
Thanks
I learned it in York Thank you Eleanor
Thank you Guy