I2PC
Instruct Image Processing Centre
JM CARAZO
Image Processing in SPA: Principles and
Workflows
José María Carazo (carazo@cnb.csic.es)
Spanish National Center for Biotechnology
Instruct Image Processing Center
Cryo-EM Course Lectures Outline
Monday
13:00-13:45
13:45-14:30
15:00-16:30
JO: General introduction to three-dimensional cryo
microscopy
SdC: Sample preparation and vitrification
MW: Modelling/Fitting
Tuesday
09:00-9:45
09:45-10:30
JMC: Single Particle Analysis (SPA): The basics
JMC: Image processing workflows for SPA
Thursday
09:00-09:45
09:45-10:30
JO: Image processing workflows for Tomography
SdC: New EM technologies
Life is based on macromolecular machines
DNA replication Protein synthesis Dynein motion
This is our objective!
The cryo-EM SPA pledge
• In 3D Electron Microscopy individual
macromolecules are visualized down to
atomic resolution.
• Trapped in ice, these molecules are free to
expose their internal flexibility/plasticity.
The cryo-EM SPA pledge
• In 3D Electron Microscopy individual macromolecules
are visualized down to atomic resolution
– Attention to every detail of the image formation
process
– Very precise image processing
• Trapped in ice, these molecules are free to expose their
internal flexibility/plasticity
– Need to classify individual images in 3D
The cryo-EM SPA pledge
Attention to every detail!
•Characterization of each image
• A posteriori characterization of the Projection
Geometry
• 3D reconstruction process
• 3D classification
• Validation
As we were saying yesterday …..
2D projections : lack of information
Limits the comprehension
of complex objects
The value of a «radiography»
Compared to full 3D CT
Compared to full 3D CT
PSF:
Point Spread Function
Object
Imaging
System
Image
Realistic Image Formation Model: CTF
characterization
Real space and Fourier Space
CTF Profile in Fourier Space
CTF Challenge
CTF Challenge
How precise should we be?
The cryo-EM SPA pledge
Attention to every detail!
• Characterization of each image
•A posteriori characterization of
the Projection Geometry
• 3D reconstruction process
• 3D classification
• Validation
Tomography Principle
Acquisition of
tilted image series
Correction of
microscope
default
(mechanical drift,
CTF...)
Reconstruction
That does not apply to SPA
Parameter space
• For each particle we need to determine 3 angles
and 2 shifts. FIVE parameters.
• If we have 100.000 particle images.
• We then have a space of 500.000 parameters!
The cryo-EM SPA pledge
Attention to every detail!
• Characterization of each image
• A posteriori characterization of the Projection
Geometry
•3D reconstruction process
• 3D classification
• Validation
Real space and Fourier Space
Principles of Fourier Reconstruction Method
More complex geometries in Fourier space
61 y 42 y
73 y
34 y
Reconstruction as a linear
set of equations
4 3
12


J
j
jjbxf
1
)()( rr
1x 2x
3x 4x


J
j
jjii xly
1
,
0,1, jil

















3
7
4
6
43
21
42
31
xx
xx
xx
xx
Project
Reproject
x

...
Iterate (n times)
ART, the “basics”
Conceptual schema
(a) (b)
(c)
ART, the “basics”
3D reconstruction approach:
• Limited number of
projections
• Image noise
• Partial lack of control over
particle homogeneity
• Particle lack of control
over data collection
geometry
Projection
images
Reconstruction
algorithms
Three-
dimensional
reconstruction
Original
specimen
The cryo-EM SPA pledge
Attention to every detail!
• Characterization of each image
• A posteriori characterization of the Projection
Geometry
• 3D reconstruction process
•3D classification
• Validation
The 3D flexibility challenge
The 3D flexibility challenge
NOW in RELION
Statistical model
Each image is a projection
of one of K underlying 3D
objects k
with addition of noise
Unknowns: the 3D objects
k, orientations
k = 1 k = 3k = 2
Parameter space of cryo EM SPA
• Target (the X’s): A volume of (for example)
100 x 100 x 100 voxels = 10**6 variables
– (Plus 500.000 = 5 x 10**5 geometry variables)
– (plus 100.000 x k (classes))
• Measurements (the Y’s): 100.000 particle
images of 100 x 100 pixels = 10**9
• But we have noise!: 2 + 2 = 5 (or 3, or 6 …)
Everything is mixed!!!
f(x)
x
local
minima
global
minimum
Parameter space of cryo EM SPA
f(x)
x
fs (x)
Parameter space of cryo EM SPA
f(x)
x
local
minima
global
minimum
The Initial Volume Problem in SPA
The cryo-EM SPA pledge
Attention to every detail!
• Characterization of each image
• A posteriori characterization of the Projection
Geometry
• 3D reconstruction process
• 3D classification
•Validation
Validation
VALIDATION:
results of our validation method on
controversial HIV data
We have applied our approach to validate the map presented in the
controversial work of MAO (1) and we have compared the results
with the map reported by Subramaniam (2)
(1) Mao Y, et al. (2013) Molecular architecture of the uncleaved HIV-1
envelope glycoprotein trimer. Proc Natl Acad Sci USA 110(39:
12428-12433
(2) Subramaniam S (2013) Structure of trimetric HIV-1 envelope
glycoproteins. Proc natl Acad Sci USA 110(E4172-E4174)
We have used the data deposited at EMDB EMPIAR 10008 (MAO) &
EMPIAR 10004 Subramaniam, for the validation we have used
approximately 1000 particles and the EMDB maps
For the validation approach we have used first the extracted
particles of MAO, which has been compared with the two maps
(MAO & Subramaniam)
The quality parameters
obtained are of
Q0 = 0.48 (Subramaniam)
Q0 = 0.13 (Mao)
For the validation approach we have used first the extracted
particles of MAO, which has been compared with the two maps
(MAO & Subramaniam)
The quality parameters
obtained are of
Q0 = 0.84 (Subramaniam)
Q0 = 0.15 (Mao)
Workflows: How do we do it in practice?
Workflows: How do we do it in practice?
Workflows: How do we do it in practice?
Using different EM software
packages is now like the
tower of Babel
Workflows: How do we do it in practice?
Importing micrographs
Preprocessing micrographs
Calculating CTF
Typical CTF displays
Particle picking
Particle picking
Typical particle picking display
Extracting particles
Sorting particles
Typical sorting display
Importing Initial Volume (and the “Problem”?)
Finding Projection Geometry
Typical Relion 3D display
Typical Volume display
The Initial Volume Problem (in the Web)
The Initial Volume Problem (in the Web)
The Initial Volume Problem (in the Web)
The Initial Volume Problem (in the Web)
Typical 3D displays
Typical 3D displays
Instruct Open call for Access
www.structuralbiology.eu
A distributed infrastructure for integrated structural biology