Automated collection and processing of macromolecular diffraction data with DNA

1

• Automated collection and processing of
  macromolecular diffraction data with DNA
• Project started in 2001 following an ESRF user
  meeting
• Currently involves scientists at Diamond, ESRF,
  EMBL Grenoble, EMBL Hamburg, SRS and SOLEIL.
• Funded (indirectly) by AUTOSTRUCT, BIOXHIT,
  eHTPX, CCP4
• Aim is to automate data collection and
  processing at synchrotron beamlines.

Data Not Analysed Definately Not
Attractive Does Not at All DogPatch
Neighbourhood Association Distributed
Network Attack Deadly Nocturnal Assassins Deer
Natural Aroma Dark Native
Apostle Drink No Alcohol Disco Night Action
Dawn of the New Age Desire and Needs
Assessment Drugz N Alchohol Douglas
Noel Adams (Thanks to acronym.com)
2
Participants

• DNA developers
• Karen Ackroyd (a)
• Alun Ashton (b)
• Gleb Bourenkov (c)
• Sandor Brockhauser (d)
• Marie-Francoise Incardona (f)
• Steve Kinder (a)
• Pierre Legrand (e)
• Karl Levik (b)
• Romeu Pieritz (f)
• Sasha Popov
• Harry Powell (g)
• Darren Spruce (f)
• Olof Svensson (f)
• Graeme Winter (a)
• DNA Exec. Committee
• Gérard Bricogne (h)
• Andrew Leslie (g)
• Sean McSweeney (f)

• Collaborating sites
• (a) CLRC Daresbury, UK
• (b) Diamond light source, UK
• (c) EMBL Hamburg, Germany
• (d) EMBL Grenoble, France
• (e) Synchrotron Soleil, France
• (f) ESRF, France
• (g) MRC LMB, Cambridge, UK
• (h) Global Phasing, Cambridge, UK
• NSLS Brookhaven

Funding from EU (AUTOSTRUCT, BIOXHIT), BBSRC
(EHTPX), CCP4
3
Current Structure Exec - policy decisions,
conflict resolution. One representative from each
institution that hosts a developer. Project
Coordinator - (Alun Ashton). Arranges VCs,
ensures actions are carried out. Developers - Do
the work Full DNA meetings twice a year.
Additional developers meetings two-three times a
year. Video conferencing every two
weeks. Bugzilla for bug tracking. CVS at D/L
(moving to Diamond)
4
Current structure of DNA 1.0
ISPyB
5
Characterising a single sample collect reference
images
Assumes that the sample is already centred in the
beam.
6

• Project Status
• DNA 1.0 released December 2004. Basically it
  didnt work.
• DNA 1.1 planned release April 2007, although
  something very close to 1.1 is already installed
  on 7 MX beamlines at ESRF.
• Improvements
• Robustness improved significantly
• Datasets integrated and scaled on the fly
• Allows a sample ranking mode for multiple
  crystals of the same type
• Significant improvement in autoindexing success
• Includes an interface to the EMBL/ESRF mini
  kappa goniostat.
• Now at a critical phase .. People are starting to
  use it !

7

• Future plans
• DNA 2.0 involves a complete rewrite of the code.
  Project manager Olof Svensson.
• Currently a spike development is underway to
  establish which tools to use.
• A scientific case is being prepared which will
  define the scientific (user) objectives.
• Objectives
• More sophisticated data collection strategies
  (multiple wavelength, multiple crystals, multiple
  sweeps)
• Feedback from downstream processing to improve
  data quality
• More sophisticated treatment of radiation damage
• Facilitate incorporation of other
  new/replacement modules, eg alternative data
  processing programs, absorption corrections etc
• Improved summary of experimental results in
  database (ISPyB)
• Automated selection of wavelength for anomalous
  data (peak).

8
The End
9
Automating the data collection step of MX - the
DNA project
Crystallisation
Phasing
Data Collection
Protein Production
Protein Structure
Target Selection
Structure analysis
Deposition
10
Objective of DNA Given some basic information
about the project and the available crystals, to
determine and carry out the best possible
diffraction experiment from the available
crystals in the available time, by a procedure
that requires little or no intervention from the
user during the experiment.
Make it easier for inexperienced users to collect
good data. Should facilitate both Fedex
crystallography and remote operation.
11

• Input to DNA
• The diffraction plan (supplied by user)
• The type of experiment (MAD/SAD phasing, sulphur
  phasing, high resolution data for refinement).
• Resolution ideal, minimum acceptable, maximum
  required.
• Type of anomalous scatterers for phasing
  experiments
• Unit cell and space group (if known)
• Crystal lifetime (if known)

• X-ray source properties (supplied by beamline
  scientists)
• Typical crystal lifetime (in seconds or photons)
• Maximum rotation rate for spindle
• Minimum safe exposure time (depends on shutter)
• Accessible wavelength range

12

• Output from DNA
• A summary of the diffraction characteristics for
  every crystal tested
• diffraction limit (based on analysis of spot
  finding/ BEST))
• unit cell, probable Laue group/lattice type
  (autoindexing)
• mosaicity (MOSFLM)

• assessment of crystal perfection (based on
  presence of multiple lattices)
• assessment of spot shape (single, streaky,
  split, multiple)

• When defined criteria have been met (resolution,
  best sample of those tested), one or more (MAD)
  data sets (h, k, l, F, sig(F)) of scaled and
  (optionally) merged structure factor amplitudes.
  The data processing may be preliminary (to save
  time) but will give a realistic estimate of data
  quality.

13
Simplified Architecture of DNA
Replace the user with an expert system, a
program with built-in decision taking. The system
needs to be modular in order to accommodate
existing site-specific beamline and sample
control software. Different data processing
software and databases should also be possible.
14
Basis of operation The expert system is
responsible for issuing requests and controlling
the sequence of operations.

• Load the next sample and centre it in the beam
  (BCM)
• Collect two initial reference images separated
  by 90o in phi at resolution stated in diffraction
  plan (BCM)
• Pre-screen images for strength of diffraction
  (if any), presence of ice rings (DPM)
• If images OK, auto-index singly and together
  (DPM - MOSFLM)

• Apply acceptance criteria to indexing (rms error
  in spot positions, age spots rejected from
  indexing, shift in direct beam position) (ES)
• If indexing successful, integrate both images
  and estimate mosaicity (DPM - MOSFLM)
• Based on results of integration, determine
  effective resolution limit of data and a data
  collection strategy (DPM - BEST).
• Write results of characterisation back to
  database (ISPyB).
• Repeat for all samples of same type, determine
  ranking (ES).
• Collect data set from the best crystal,
  integrate data as it is collected, run quick
  scaling (SCALA).

15
Implementation of DNA at ESRF DNA is installed on
all MX beamlines except ID13. Started with a
desktop icon (after starting MxCube)

• Currently two modes of operation
• Characterising a single sample
• Sample screening (pipeline mode)

16
Characterising a single sample index reference
images
17
Characterising a single sample calculate strategy
18
Characterising a single sample modify strategy
19
Characterising a single sample examine results I
20
Characterising a single sample examine results
II
21
Characterising a single sample examine results
III
22
Characterising a single sample examine results IV
23
Characterising a single sample collect data
24
Characterising a single sample refining cell
prior to integration
25
Characterising a single sample integrating and
scaling data
26
Characterising a single sample examining log
files (SCALA)
27
Sample screening get list of samples
28
Sample screening select samples
29
Sample screening ranking
30
Sample screening ranking criteria
31
Sample screening ranking by resolution
32
View Rank Result allows a quick comparison of
samples
33
Select one or several samples for data collection
If several samples are chosen, only automatic
collection is allowed (no opportunity to modify
data collection parameters).
34
DNA screening / data collection scenarios
Process
Integration, strategy and ranking
Index 1st image Index 2nd image Index both images
Screening/ ranking scenario
Pre- screen
Change sample
Collect two images 90 apart
Time
3 mins
35

• Issues
• Conceptual
• How ambitious to be in terms of handling less
  than perfect crystals.
• How much user control should be accommodated ?
• Technical
• Depends on success of crystal centring
• Dependency of success of autoindexing
• Choice of criteria for ranking
• How to deal with user requested resolution
• How to deal with radiation damage
• How to make processing keep up with data
  collection

36

• Future Plans
• Scrap it all and start again
• Extend to include collection and analysis of
  edge scans for design of anomalous diffraction
  data collection
• Optimise beam size wrt crystal size
• Allow more sophisticated data collection
  strategies (about more than one axis, variable
  exposure time/oscillation angle per image, using
  more than one position on a crystal or several
  different crystals)
• Allow strategy to be updated if initial estimate
  of point group was incorrect
• Automatic determination of crystal lifetime from
  a sacrificial crystal.
• Improve level of information stored in the ISPyB
  database
• Improve robustness

37
Diffraction plan for ISPyB