POINTLESS &
SCALA
Phil Evans
POINTLESS
What does it do?
1. Determination of Laue group & space group from unmerged data
i. Finds highest symmetry lattice consistent with unit cell
ii. Scores each potential rotational symmetry operator in lattice, using correlation
coefficient on normalised intensities E2, with a derived “probability”
iii. Scores all combinations of symmetry operators to derive a probability for each
point group which is a sub-group of the lattice group (hence find “Laue group” = point
group + lattice centering)
iv. Scores potential systematic absences to detect screw axes, using a Fourier
analysis of I/σ(I), hence assign a probability to possible space groups, if possible
1. In cases where there are alternative indexing conventions, match the indexing to a
reference file (merged or unmerged)
2. Reindex or change space group (cf program REINDEX)
3. Just sort one or more input files for Scala (cf SORTMTZ)
POINTLESS
New features
• Input from XDS & Scalepack formats
XDS_ASCII.HKL or INTEGRATE.HKL
Note the Scalepack output is still not suitable for
proper scaling in Scala, since geometric information is
lost.
• Multiple input files, explicit or with wild-cards
Checks for consistent indexing between files or file
series
★ Automatic renumbering of batches to make them unique
(a long-standing irritation)
★
• Defaults to IUCr standard settings: can be overridden
a < b < c for primitive orthorhombic, allows eg P 21 2 21
★ I2 in place of C2 to give smallest β angle
★
POINTLESS
Improved scoring schemes
1. Probability scoring uses Lorentzian distribution (larger
tails than Gaussian)
2. Systematic absence
scoring uses intensities
“corrected” by
subtraction of small
fraction (0.02) of their
neighbour, to allow for
very strong reflections
bleeding into absent
neighbours. Most
reflections are
unaffected.
Fourier I/sig(I)
Corrected I’/sig(I)
term
Peak
Probability Peak
Probability
63
0.624
0.055
0.856
0.038
62
0.695
0.068
0.872
0.039
61
0.684
0.248
0.720
* 0.703
Correction gives higher peaks, larger
probabilities
CCP4i interface
POINTLESS
General options
Multiple file
input,
same
dataset
Options for setting
SCALEPACK example
POINTLESS
Scalepack files do
not include the unit
cell, so this must be
given
Result displayed
by Baubles
POINTLESS
Consistent indexing to reference file
(merged or unmerged)
Spacegroup H3
SCALA
SCALA
New developments
• Corner correction to apply externally calculated
correction table as function of detector position
Correction
table for ESRF
ID23-1
Generated
from many
diffraction
patterns
(from Chris
Nielsen
et al.)correction ≈ 1.4 !
Maximum
SCALA
•Automatic optimisation of SD correction parameters
Before
After
Optimisation of σ’2 = SDfac2 [σ2 + SdB <Ih> + (SdAdd <Ih>)2]
Minimises deviation of Sigma(scatter/σ) from 1
ie flattens out the plot
Uses simplex minimisation (thanks to Kevin for code)
SCALA
Minor things
• Changed logfile to contain Results section for Baubles
• Resolution limits for different datasets (in addition to limits by run)
• Output of multiple datasets to same file: could go into Truncate at the same time
(OUTPUT AVERAGE TOGETHER)
• ROGUEPLOT to plot outliers on detector
Outliers along
rotation axis
Clusters of outliers
around tile corners
SCALA
Summary from Baubles
Summary table (aka Table 1)
LogGraphs
Future developments
Pointless & Scala: no major developments planned
Bug fixes, respond to complaints, a few small things to change
New program, working title AIMLESS
Probably eventually all part of same program with
POINTLESS, rather than a separate one: name to be chosen!
Essentially a rewrite of Scala: work has begun
Advantages:
• More flexibility
• Possible new scale models
• time extrapolation
• detector corrections, cf XDS MODULATION correction
• Better analysis etc.
•Assessment of data, advice for user: automatic optimisation
of resolution limits, radiation damage vs. completeness, etc,
etc
Acknowledgements
Ralf Grosse-Kunstleve
cctbx
Kevin Cowtan
clipper, simplex, C++
advice
Martyn Winn & CCP4 gang ccp4 libraries
Peter Briggs
ccp4i
Airlie McCoy
C++ advice, code etc