New Scientific Opportunities through Intelligent Automation and High-Throughput Crystallography

1
New Scientific Opportunities through Intelligent
Automation and High-Throughput Crystallography
The APS Users Meeting 2000
S. Michael Soltis, Stanford Synchrotron Radiation
Laboratory, soltis_at_slac.stanford.edu SSRL is
funded by the US Dept. of Energy and the National
Institutes of Health
2
Scientific Challenges and Opportunities
Abstract Synchrotron beam lines have become
extremely productive and efficient structure
determination facilities. Advance instrumentation
such as large-area, high-speed readout CCD
detectors and automated beam conditioning
systems, coupled with intuitive and reliable
software interfaces within a high-performance
computational environment, are maximizing user
throughput while increasing the experimental
control and flexibility for advanced data
collection strategies. The real-time feedback of
experimental results from the data collection
allows for optimized data collection strategies
and enables phasing and structure determination
evaluation during the experimental run. By making
advances in data collection and structure
determination methods and techniques, new
opportunities can be realized (1) large and
complex systems become more amenable to
high-resolution structure determination, (2)
high-throughput screening methods may
reinvigorate structure aided drug design and (3)
high-throughput structure determination will make
structural genomics a reality.

• The Next Generation
• Automation and robotics
• 3 x 3 CCD
• High-Throughput Structure Determination
• Science
• Large assemblies and molecular machines
• Poorly diffracting crystals
• Drug targets
• Structural genomics

• Current State of the Art
• Developments
• BL9-2
• BLU-ICE
• High-end computing
• MAD Kr
• Ultrahigh
• Direct methods
• Auto-tracing
• Scientific Outcome
• Yeast RNA polymerase II
• Ultra high resolution structures
• Bucandin
• Calmodulin

3
Synchrotron Radiation in Structural Molecular
Biology

• Small Angle Diffraction/Scattering
• protein folding
• oligomer assembly
• conformational changes
• low resolution virus structures
• X-ray Absorption Spectroscopy
• analysis of metal centers in proteins
• reaction intermediates in metalloproteins
• solution vs. crystalline state of active site
  conformation
• Macromolecular Crystallography

BL 4-2 SAXS/SAXD
BL 11-1 SU/TSRI/SSRL Monochromatic
BL 9-1 Monochromatic
BL 1-5 Multi-wavelength

• BL 9-2 Multi-wavelength

BL 7-1 Monochromatic
BL 7-3 Biological XAS
BL 9-3 Biological XAS
4
BL9-2 - Optimized for MAD with a High Degree of
Automation

• Optical Components
• Table
• Sample Alignment
• Kappa
• x,y,z of sample
• Align outside the hutch
• Beam Stop
• Up to 250 mm
• Simple to align
• Quantum-4 CCD Positioner
• Detector distance (70-800 mm)
• Horizontal and vertical
• Tilt in vertical
• Fluorescence Detector
• Linear position
• Beam Conditioning
• Variable beam size
• Variable Attenuation
• Guard shield

5
Beam Conditioning System
6
Distributed Software Architecture

• Distributed Devices and Systems
• detector
• goniometer
• shutter
• beam monitors
• filters
• slits
• motors
• encoders
• video
• microscopes
• fluorescence
• cryostat
• Distributed Control System Server (DCSS)
• Cross-platform
• Accommodates a wide range of hardware and control
  systems
• Simple to expand
• Multi-session interfaces
• Private Network for Secure and Reliable
  Communication

7
Beam Line Unification and Interface Control
Environment (BLU-ICE)A Graphical User Interface
8
Collect Tab
9
Energy Scan Tab
10
Staff Tab
11
Current Computational Environment

• Real Time Feedback
• High performance computers and multiprocessor
  servers
• Fast reliable networks and switches
• 4 X 600 MHz CPU ALPHA Server
• Image transformation
• Data processing software
• Disk Space
• 750 GB
• Dual Processor SGI Octanes with 2 Monitors
• Single interface
• Local processing and backup
• Excellent graphics capability
• Stereo
• Redundant Backup Systems
• 4mm, 8mm, DLT, CD, (DVD)
• Windows NT Server for PC applications
• WEB Server - Tools and Documentation

12
Beam Line Setup
13
Collaboratory for Protein Crystallography
THE NATIONAL UNIVERSITYofSINGAPORE

• Facilitate World Wide Collaborations
• Remote Collaborators Participate in Data
  Collection through Publication
• Relatively Novice Users Perform Challenging
  Experiments
• Utilize and Integrate National Resources
• Synchrotron beam lines
• 1000 TB storage and retrieval capability
• High-performance compute farms

14
National Resources Collaborate to Increase
Efficiency and Effectiveness
Data Reduction and Structure Analysis
15
A Beam Line Standard

• Goals Achieved
• Efficient and productive data collection/structure
  determination facility
• Typical time for novice users to start collecting
  data - 20 minutes
• Streamlined MAD data collection
• 4 MAD structures solved on site (BL9-2) during
  the last 2 months
• Beam conditioning was critical for quality data
  collection in several cases
• Time to reconfigure beam line focus or energy
  range - 15 minutes
• Upgrade BL1-5, BL7-1, BL9-1 - BL11-1
• Portability of DCSS and BLU-ICE
• Yeast RNA Polymerase II (R. Kornberg Laboratory,
  Stanford)

• Transcription of DNA into RNA - key step in gene
  expression underlying all aspects of cellular
  metabolism
• 450 kDa complex of 10 subunits
• 10 years of data collection and refinement of
  crystallization and cryo-cooling conditions
• BL9-2 accelerated screening
• Solved to 3 Å
• Recently identified 80 SeMet positions (50)
• Native data collected to 2.8 Å resolution

P. Cramer, et al. Science, 288, 640 (2000)
16
Enabling New Science

• Through Automated Crystal Screening

Automated Data Collection/Processing and Improved
Phasing Methods
Structural Genomics
17
Automated Sample Handling and Characterization

• Cassettes
• Holds 96 flash-cooled samples
• Designed for commercial shipping Dewars
• Keyed for tracking samples
• Mounting Samples on the Diffractometer
• Commercial robot
• Dispenser box
• Transfer at liquid nitrogen temperature

• Screening Software
• Auto-alignment of sample
• Determine quality and strategy
• Screen for anomalous signal

18

• Co-organizers
• Paul Phizackerley and Ashley Deacon (SSRL) -
  sample cassette and mounting system using a
  commercial robot
• Thomas Earnest (LBNL / ALS) - The role and
  direction of automation for future structural
  biology research
• Gerd Rosenbaum (SER-CAT APS/ UGA) - overview
• Speakers
• Hassan Belrhali (ESRF) - sample changer
• Malcolm Capel (BNL) - beam visualization system
  and robotic crystal mounter
• Florent Cipriani (EMBL-Grenoble Outstation)
• Hassan Belrhali (ESRF) - Micro-diffractometer
• Earl Cornell (LBNL) - automated viewing and
  centering of protein crystals
• Steven Muchmore (Abbott Laboratories) - an
  automated system for mounting, aligning and
  collecting diffraction data
• Deming Shu (ANL) - preliminary design of an
  automated crystal mounting device
• Derek Yegian (LBNL) - task-specific robotics for
  sample loading, centering and retrieval
• 60 participants

19
Large Area and Rapid Read-Out Detectors

• ADSC Quantum-315 for BL9-2 and BL11-1
• Readout 10 times faster than Quantum-4
• 1 - 0.3 sec readout
• High throughput data collection
• Completeness and redundancy
• Fine phi slicing
• Reduced time dependent radiation damage
• 315 mm X 315 mm active area
• Ultra-high resolution
• Large unit cells
• Improved profiling
• 8 CPU ALPHA Server and 3 TB RAID
• 80 MB/image/sec
• Parallel image processing
• Pixel Array Detectors
• Fast readout
• Large size

Quantum-315 CCD Detector
20
Kr MAD

• Myoglobin
• Known Xe binding site
• Incubate 2 min
• 400 PSI Kr gas
• 3 wavelengths
• 1.7 Å resolution
• 99.6 completeness
• 20 fold redundancy
• Rmerge 0.091
• I/? 5.3
• Kr Peak 19.4 ?
• 63 Occupancy
• Phased with SHARP
• FOM (acentric reflections) 0.87
• FOM (centric reflections 0.73
• General Applicability
• Binding in 50 proteins
• Accessible energy (Kedge 14.3 KeV , 0.87 Å)

21
Ultra-High Data Collection

• BL9-1
• MAR345
• ? 0.70 Å
• Multiple Passes
• Protect Detector
• Small Oscillations
• Completeness
• Redundancy

• Standard Ultra-high mode on 9-1
• 15 Active Proposals

• BL9-2
• ? 1.0 - 0.6 Å
• Quantum-4 CCD
• Offset Detector
• Multiple Passes
• Small Oscillations

22
Structures at Sub-Ångstrom Resolution
PYP at 0.85 Å
Neuraminidase at 0.85 Å
PYP - U.K. Genick et al. Nature 392, 206 (1998)
Neuraminidase - E. Garman (Oxford University),
T. Schneider (University of Goettingen), G. Laver
(Australian National University) Subtilisin -
Kuhn et al. Biochemistry 37, 13446 (1998) B-DNA -
C. L. Kielkopf et al. JMB 296, 787 (2000)
23
A Novel Three-Finger Toxin Bucandin

• Synaptic neurotoxin
• Data collected to 0.95 Å resolution on BL9-1
• Ab-initio structure determination with both Shake
  and Bake (SnB) and SHELXD (half-baked), using
  data to 1.1 Å resolution
• 61 of 63 amino acids modeled using wARP
  (including side chains)
• Solved and traced within 24 hr using
  multiprocessor ALPHA

Map after refinement
Map from SnB
A. Deacon (while at CHESS), P. Kolatkar
(University of Singapore), I. Usón (University of
Göttingen, Germany), SMB Staff
24
Calmodulin Determined by Direct Methods

• Data collected to 1.1 Å resolution on BL9-2 (A.
  Deacon, M. Wilson and A. Brünger)
• Structure determined ab-initio by SnB in less
  than 1 hour
• Largest structure solved by SnB to date
• 150 amino acids
• 1192 protein atoms
• 5 calcium atoms
• 160 bound water molecules
• wARP - automatic model building
• 93 of main chain and 65 of side chains
• 4 hours on ALPHA

25
Arsenite Oxidase

• Long chain contains Mo site and a 3Fe/4S cluster
• Long chain is homologous to other members of the
  DMSO reductase family, but no protein side chain
  is bound to Mo
• Short chain contains a Rieske-type 2Fe/2S site
• Short chain has no counterpart among known Mo
  protein structures but is similar to other Rieske
  Fe/S protein domains

• 100 kDa monomer, dimer in the asymetric unit
• Hg Mo SIRAS, Fe MAD
• Phased with SHARP
• wARP auto-traced 1800 of 2000 amino acids to 1.8
  Å
• Electron density for sequencing of 95 of all
  amino acids

26
Automated Data Processing and Structure
Determination

• Automated Data Processing
• Development of methods and algorithms will be
  guided by our collaborators
• Scripts
• Structure Solution
• Se, Kr, other
• Auto-Tracing and Auto-Refinement
• ARP/wARP (perhaps 2.5 Å resolution limit)
• Non-crystallographic symmetry and prosthetic
  groups
• CNS, XtalView, CCP4 (advanced tracing and model
  building tools)

27
Center for Structural Genomics
The Salk Institute
28
SSRL is Funded by the US Dept. of Energy (BES,
BER) and the NIH (NCRR, NIGMS)

• SSRL Director - Keith Hodgson
• SMB Leader - Britt Hedman
• The Macromolecular Crystallography Group

Mike Soltis Peter Kuhn Henry Bellamy Aina
Cohen Ashley Deacon Paul Ellis Thomas
Eriksson Ana Gonzales Mike Hollenbeck Scott
McPhillips Tim McPhillips Pavel Petrashen Paul
Phizackerley Amanda Prado Jere Rassai