Automated Data Collection for CryoTEM

1
Automated Data Collection for CryoTEM

• Clint Potter
• Imaging Technology Group
• Beckman Institute
• University of Illinois at Urbana-Champaign

2

• Collaborators
• University of Illinois Bridget Carragher, Jim
  Pulokas,
• Nick Kisseberth, Amy Reilein, Carmen Green,
  Nebojsa Jojic
• National Institutes of Health Jenny Hinshaw
• Scripps Research Institute Ron Milligan
• Support
• National Science Foundation (DBI-9730056)
• IBM Sponsored University Research Program
• Informix, Inc.
• Beckman Institute

3
Outline
1. Overview 2. Methods emScope Software
control of TEM Goniometer positioning
accuracy 3. Applications of the Leginon
system Catalase project Cryo-TEM
project 4. Conclusions and next steps
4
Automated Data Collection for CryoTEM

• Goal Completely unattended acquisition of very
  large numbers of high quality micrographs.
• Motivation 1000s of images necessary to solve
  structures at high resolution.

6,600x
38,000x
660x
5
What Needs to be Automated?

• Specimen preparation, handling, loading
• Feature selection
• Focus, Astigmatism correction
• Image acquisition ( film, digital)
• Quality assessment
• Scope maintenance (cryogen refills, film
  changes, etc.)

6
emScope- Software Control of TEM

• Portable and extensible toolkit for control of a
  TEM and camera.
• Layered software architecture
• 6 Layers from low level instrument control
  through graphical interface.
• Distributed application support.
• Support for Tcl/Tk and JAVA languages.
• One library supports many instruments (CM
  Series, EM-430, Gatan MSC)
• Kisseberth, JSB, 120, 309-319 (1997)

7
Applications of the emScope Toolkit

• GridScan

EMact
8
Automation Goniometer Positioning Accuracy

• Problem Accurately move stage to predicted
  location on the image.
• Error is 10 of distance moved.
• For a 10 mm move

10 mm
660x
Predicted 1 mm error Location
- FOV at 38,000x.
9
Automation Goniometer Positioning Accuracy

• Problem Accurately move stage to predicted
  location on the image.
• Error is 10 of distance moved.
• For a 10 mm move

Accurate relocation
1 um error
- FOV at 38,000x on 1Kx1K camera.
10 mm
660x
10
Characterization of Slew Rate

• Measured piecewise slew rate (nm/tick) over range
  of goniometer (CompuStage) using image cross
  correlation.
• Results 18 periodic variation over range of
  goniometer.
• Slew rate for X and Y axis a function of position

Y axis (period 41.6 mm)
X axis (period 61.9 mm)
11
Goniometer Modeling Results

• 2 CompuStages characterized with similar results.
• Slew rate for X and Y modeled with Fourier Series
• Validation RMS error as a function of distance
  moved over range of goniometer

RMS Error (mm)
Distance Moved (mm)
12
Goniometer Modeling Conclusions

• Modeling slew rate provides a 7x increase in
  goniometer accuracy with no dependence on
  distance moved.
• Accurately position goniometer to within 100 nm.
• Characterization of goniometer is automated and
  can be completed in a few hours.
• Technique can easily be applied to other
  goniometers.
• Stable with no drift over months.
• Characterization is a QC maintenance tool.

13
Leginon System Architecture
Microscope Philips CM200
Leginon
Microscope Server IBM RS6000
Low mag acquire
Microscope Application Client
High mag acquire
emScope
Data assessment
SGI or IBM RS6000
Gatan MSC camera
14
Leginon System Architecture
emScope Library
Tcl/Tk Extension
Leginon
Microscope Philips CM200
GridScan
Microscope Server
Command Line Applications
JavaScope Server
JavaScope Client
Tomography/EMact
Gatan MSC Apple PowerMac/G3
Client emScope Application
15
Applications of the Leginon System

• Catalase Project
• Identify good crystals.
• Cryo-TEM
• Identify good holes.

16
Use of Leginon in the Catalase Project
High mag - 38000x For each identified feature
Autofocus Capture image Assess quality of
image
Low mag - 660x For each grid square Capture
image Identify feature of interest Locate
feature at center
17
Automated Identification of Features of Interest
660x
18
Automated Identification of Features of Interest
Overlapping areas can be separated using a
mixed-model algorithm
19
Catalase Project Results Data Acquisition
Efficiency

• Acquire 1 high magnification image every 90
  seconds (1000 images/day)
• Bottlenecks
• - CPU, networks, algorithms,
• software implementation, etc.

20
Human Image Quality Assessment
Qh - Human rated image quality - Number of
layer lines.
Power spectrum
High mag. image
21
Machine Image Quality Assessment
Qm - Machine rated image quality - Number
of diffraction spots with S/N gt 3.5
Diffraction spots
Power spectrum
22
Machine vs. Human Image Quality Assessment Metric
Performance
Qm
Qh
n 1681
23
Catalase Project Performance Evaluation
Qm65 n Human operator 79
288 Fully-automated 51 380 Intelligent
86 115
T
Quality
Feature Intensity
24

Use of Leginon in the Cryo-TEM Project
Low mag. - 660x For each grid square Capture
image Identify feature of interest Locate
feature at center
High mag. - 38000x For each identified feature
Autofocus Capture image Assess quality of
image
25
Quantifoil Grids

• Perforated support foils with pre-defined hole
  size, shape and arrangement (Ermantrout et al.,
  Ultramicroscopy 74)

Hole diameter 2 mm Hole spacing 4 mm
26
Leginon for Cryo-TEM Process Overview
1. Low mag. image of grid square (660x) 2. Find
good holes based on a) Quantifoil lattice
parameters hole size, spacing, geometry b) Ice
thickness k ln ( Ibf / Iice) Euseman,
1982 For each good hole 3. Medium mag. image
of hole (6600x) 4. High mag. image of hole
(38000x)

660x
6600x
38000x
27
Leginon Graphical User Interface
28
Leginon for Cryo-TEM Good hole identification
Ice thickness k ln ( Ibf / Iice) Euseman,
1982
29
Leginon for Cryo-TEM Rip detector

• Automatically identify torn grid squares

30
Results Acquisition Efficiency
Mar03 Experiment - microtubules - 12 hour fully
automated acquisition. - 200 grid squares
evaluated. - 70 low magnification images
analyzed. - 444 target holes identified 443
good holes, 1 miss - Image acquisition rate
115s/image.

31
Results Automated Focusing and Astigmatism
Correction

• Automated Focus
• Measured Defocus Defocus Shift n
• Carbon 1.8 /-0.1 um 0.04 /-1.4
  um 200
• Ice 1.3 /-0.2 um 0.03 /-3.5 um 70
• Automated
• Astigmatism Correction

32
Automated Data Collection for Cryo TEM
Conclusions

• Demonstrated full automation of
• Feature selection
• Focus, Astigmatism correction
• Image acquisition ( film, digital)
• Quality assessment

33

• Changes to the instrument
• Extended operation (cold trap, cold stage)
• Fine control of goniometer
• Low dose kit stable at low magnification
• Automated aperture insertion and alignment
• Larger capacity film cassette

Modified LN2 dewar on Philips CM200
34
Next Steps

• 1. Technology Transfer
• Milligan laboratory at Scripps
• 2. Holey carbon grid

35
Next Steps
3. 6,600x Image Analysis

6,600 x
38,000 x