stanford-slac cryo-electron microscopy - dealing with … · 2020. 8. 12. · fourier transform...
TRANSCRIPT
-
Dealing with Resolution Anisotropy andConsiderations for High Resolution Structure
Determination
Dmitry LyumkisSalk Institute for Biological Studies
American Crystallographic Association, cryo-EM Workshop, August 11, 2020
-
• Preferred orientation — the origin of resolution anisotropy
Outline
• Resolution anisotropy in cryo-EM reconstructions (caused
by preferred orientation) and its evaluation with 3D FSCs
• Combatting anisotropy using tilted data collection strategies
• The (unexpected) effect of sampling (in)homogeneity on
global resolution
• Evaluating sampling (in)homogeneity using the SCF criterion
• Considerations for high-resolution
-
Origin of resolution anisotropy in cryo-EM:The idealized behavior of particles on a grid
Noble, A. J., et al. (2018). Routine single particle CryoEM sample and grid characterization bytomography. eLife, 7, 32.
-
Noble, A. J., et al. (2018). Routine single particle CryoEM sample and grid characterization bytomography. eLife, 7, 32.
Origin of resolution anisotropy in cryo-EM:The reality … Particles adhere at the air-water interface
-
Origin of resolution anisotropy in cryo-EM:The reality … for virtually all samples
Noble, A. J., et al. (2018). Routine single particle CryoEM sample and grid characterization bytomography. eLife, 7, 32.
-
• Preferred orientation — the origin of resolution anisotropy
Outline
• Resolution anisotropy in cryo-EM reconstructions (caused
by preferred orientation) and its evaluation with 3D FSCs
• Combatting anisotropy using tilted data collection strategies
• The (unexpected) effect of sampling (in)homogeneity on
global resolution
• Evaluating sampling (in)homogeneity using the SCF criterion
• Considerations for high-resolution
-
90°
Interaction at the air-water interface results in:“preferred particle orientation” — Hemagglutinin (HA) trimer
Tan, Y. Z. et al. (2017). Nature Methods, 14(8), 793–796.
-
Interaction at the air-water interface results in:smearing and elongation along Z-axis (loss of Z resolution)
Tan, Y. Z. et al. (2017). Nature Methods, 14(8), 793–796.
beamdirection
beamdirection
-
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
Fourier Shell Correlation (FSC): a quantitative way toevaluate resolution in cryo-EM
● Calculate the FSC between two half maps as afunction of spatial frequency over the entireFourier shell
Harauz, G., & van Heel, M. (1986). Exact filters for general geometry 3-dimensionalreconstruction. Optik, 73(146), 146–156.
-
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
● Calculate the FSC between two half maps as afunction of spatial frequency over the entireFourier shell
Harauz, G., & van Heel, M. (1986). Exact filters for general geometry 3-dimensionalreconstruction. Optik, 73(146), 146–156.
Fourier Shell Correlation (FSC): a quantitative way toevaluate resolution in cryo-EM
-
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
● Calculate the FSC between two half maps as afunction of spatial frequency over the entireFourier shell
Harauz, G., & van Heel, M. (1986). Exact filters for general geometry 3-dimensionalreconstruction. Optik, 73(146), 146–156.
Fourier Shell Correlation (FSC): a quantitative way toevaluate resolution in cryo-EM
-
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
● Calculate the FSC between two half maps as afunction of spatial frequency over the entireFourier shell
Harauz, G., & van Heel, M. (1986). Exact filters for general geometry 3-dimensionalreconstruction. Optik, 73(146), 146–156.
Fourier Shell Correlation (FSC): a quantitative way toevaluate resolution in cryo-EM
-
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
● Calculate the FSC between two half maps as afunction of spatial frequency over the entireFourier shell
Harauz, G., & van Heel, M. (1986). Exact filters for general geometry 3-dimensionalreconstruction. Optik, 73(146), 146–156.
Fourier Shell Correlation (FSC): a quantitative way toevaluate resolution in cryo-EM
-
● Calculate the FSC between two half maps as afunction of spatial frequency over the entireFourier shell
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
Harauz, G., & van Heel, M. (1986). Exact filters for general geometry 3-dimensionalreconstruction. Optik, 73(146), 146–156.
Fourier Shell Correlation (FSC): a quantitative way toevaluate resolution in cryo-EM
-
● Calculate the FSC between two half maps as afunction of spatial frequency over the entireFourier shell
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
Harauz, G., & van Heel, M. (1986). Exact filters for general geometry 3-dimensionalreconstruction. Optik, 73(146), 146–156.
FSC 0.143
Fourier Shell Correlation (FSC): a quantitative way toevaluate resolution in cryo-EM
-
● Calculate the FSC between two half maps as afunction of the direction (and looking atcross-correlations over a local cone of dataand not the entire shell)
3D FSCs: a quantitative way to evaluate directional resolutionand anisotropy in single-particle cryo-EM
Tan, Y. Z. …, Lyumkis D. (2017). Nature Methods, 14(8), 793–796.
-
● Calculate the FSC between two half maps as afunction of the direction (and looking atcross-correlations over a local cone of dataand not the entire shell)
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
Tan, Y. Z. …, Lyumkis D. (2017). Nature Methods, 14(8), 793–796.
3D FSCs: a quantitative way to evaluate directional resolutionand anisotropy in single-particle cryo-EM
-
● Calculate the FSC between two half maps as afunction of the direction (and looking atcross-correlations over a local cone of dataand not the entire shell)
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
FSC at View #1
Tan, Y. Z. …, Lyumkis D. (2017). Nature Methods, 14(8), 793–796.
3D FSCs: a quantitative way to evaluate directional resolutionand anisotropy in single-particle cryo-EM
-
● Calculate the FSC between two half maps as afunction of the direction (and looking atcross-correlations over a local cone of dataand not the entire shell)
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
FSC at View #2
Tan, Y. Z. …, Lyumkis D. (2017). Nature Methods, 14(8), 793–796.
3D FSCs: a quantitative way to evaluate directional resolutionand anisotropy in single-particle cryo-EM
-
● Calculate the FSC between two half maps as afunction of the direction (and looking atcross-correlations over a local cone of dataand not the entire shell)
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Spatial Frequency
Four
ier
Shel
lC
orre
lati
on
FSC at View #3
Tan, Y. Z. …, Lyumkis D. (2017). Nature Methods, 14(8), 793–796.
3D FSCs: a quantitative way to evaluate directional resolutionand anisotropy in single-particle cryo-EM
-
● Calculate the FSC between two half maps as afunction of the direction (and looking atcross-correlations over a local cone of dataand not the entire shell)
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Conical FSCsAcross Entire
Sphere
Tan, Y. Z. …, Lyumkis D. (2017). Nature Methods, 14(8), 793–796.
3D FSCs: a quantitative way to evaluate directional resolutionand anisotropy in single-particle cryo-EM
-
● Calculate the FSC between two half maps as afunction of the direction (and looking atcross-correlations over a local cone of dataand not the entire shell)
Fou
rier
Tra
nsf
orm
Hal
f Map
1H
alf M
ap 2
Conical FSCsAcross Entire
Sphere
3D FSC
Tan, Y. Z. et al. (2017). Nature Methods, 14(8), 793–796.
3D FSCs: a quantitative way to evaluate directional resolutionand anisotropy in single-particle cryo-EM
-
Evaluating resolution anisotropy: the 3D FSC
Tan et al. (2017). Nat Methods, 14(8), 793-796. Aiyer et al. in press. MiMB.
beamdirectio
n
beamdirectio
n
beamdirectio
n
Anis
otro
pic
case
Isot
ropi
c ca
se
-
Implementation of 3D FSCs in Chimera
Aiyer et al. in press. MiMB.
-
Web-server for remote 3D FSC processing:https://3dfsc.salk.edu
www.github.com/LyumkisLab/3DFSC
http://www.github.com/LyumkisLab/3DFSC
-
• Preferred orientation — the origin of resolution anisotropy
Outline
• Resolution anisotropy in cryo-EM reconstructions (caused
by preferred orientation) and its evaluation with 3D FSCs
• Combatting anisotropy using tilted data collection strategies
• The (unexpected) effect of sampling (in)homogeneity on
global resolution
• Evaluating sampling (in)homogeneity using the SCF criterion
• Considerations for high-resolution
-
Tilting the grid in the microscope helps to alleviateresolution anisotropy caused by preferred specimen
orientation
Lyumkis, D. (2019). JBC, 294(13), 5181–5197.
Tan et al. (2017). Nat Methods, 14(8), 793-796.
-
Why do tilts work?Filling in more Fourier space density (better Fourier sampling)
reconstruct
reconstruct
-
Tilting ameliorates resolution anisotropy and improves maps
Unt
ilted
Tilte
d
-
Anisotropic reconstructionuntilted data
(more) isotropic reconstructiontilted data
Passos, Li et al. (2020). Science
Tilting ameliorates resolution anisotropy and improves mapsHIV-1 intasome complexes (from Thursday ACA session)
-
Examples of tilted data collection strategies in the literature
Zhang, Y., … Shi, Y. (2020). bioRxiv, 9,2020.03.27.009381
~100 MDa Nuclear Pore Complex
Herzik, M. A., Wu, M., & Lander, G. C. (2019).Nature Communications, 10(1), 1–9.
~43 kDa Protein Kinase A
-
Possible limitations associated with tilted datacollection strategies
1. Defocus gradient across the micrograph
2. More beam-induced movement
3. Thicker ice
-
Compensating for limitations of tilts usingcomputational image processing: AAV
Working hypothesis: most of the problems associated with tilts couldbe ameliorated using computational strategies
0°st
age
tilt
60°s
tage
tilt
-
• Preferred orientation — the origin of resolution anisotropy
Outline
• Resolution anisotropy in cryo-EM reconstructions (caused
by preferred orientation) and its evaluation with 3D FSCs
• Combatting anisotropy using tilted data collection strategies
• The (unexpected) effect of sampling (in)homogeneity on
global resolution
• Evaluating sampling (in)homogeneity using the SCF criterion
• Considerations for high-resolution
-
75K particles
Passos, D. O., & Lyumkis, D. (2015). Journal of Structural Biology, 192(2), 235–244.
High-resolution dataset, with multiple “preferredorientations” for characterizing the effects of sampling
-
75K particles 10K particles
10K particlesorientation 1
10K particlesorientation 3
10K particlesorientation 2
Dataset could be subdivided into four groups by orientation
-
Reconstructions differing only in orientation distributionshave distinct SSNR profiles
-
• Preferred orientation — the origin of resolution anisotropy
Outline
• Resolution anisotropy in cryo-EM reconstructions (caused
by preferred orientation) and its evaluation with 3D FSCs
• Combatting anisotropy using tilted data collection strategies
• The (unexpected) effect of sampling (in)homogeneity on
global resolution
• Evaluating sampling (in)homogeneity using the SCF criterion
• Considerations for high-resolution
-
Dataset signal power /noise power at different
spatial frequencies
Average numberof times lattice
point k is sampled
Geometricalfactor
Per-particle SSNR
Heymann, J. B., IUCr. (2019). Single-particle reconstruction statistics: a diagnostic tool in solving biomolecularstructures by cryo-EM. Acta Crystallographica. Section F, Structural Biology and Crystallization Communications,
75(1), 33–44.
Decoupling sampling with noise variance
Derivation decouples a sampling factor from the per-particle SSNR
Previously …
-
Baldwin, P. R., & Lyumkis, D. (2019). Progress in Biophysics and Molecular Biology.http://doi.org/10.1016/j.pbiomolbio.2019.09.002
Evaluating SCF behavior using different Fourier samplingschemes
-
Decrement of the SCF for side-like views: (fully sampled)
-
Decrement of the SCF for top-like views:(Incompletely sampled)
45-degree cone,10% sprinkles
45-degree cone,3% sprinkles
45-degree cone,1% sprinkles
45-degree cone,0% sprinkles
Uniform
-
Knowing the SCF can predict the attenuation of the SSNR:side-like distributions, varying phi-angle modulation
-
Knowing the SCF can predict the attenuation of the SSNR:side-like distributions, varying phi-angle modulation
-
Knowing the SCF can predict the attenuation of the SSNR:top-like distributions, varying cone size with 3% “sprinkles”
-
Knowing the SCF can predict the attenuation of the SSNR:top-like distributions, varying cone size with 3% “sprinkles”
-
GUI for calculating SCF values:Example with top-like distribution, untilted and tilted
Baldwin, P. R., & Lyumkis, D. (2020).bioRxiv, 3(1), 2020.06.08.140863.
www.github.com/LyumkisLab/SamplingGui
Baldwin, P. R., & Lyumkis, D. (2020).PBMB, Jul 6:S0079-6107(20)30060-2
SCF* = 0.02
SCF* = 0.49
http://www.github.com/LyumkisLab/SamplingGui
-
• Preferred orientation — the origin of resolution anisotropy
Outline
• Resolution anisotropy in cryo-EM reconstructions (caused
by preferred orientation) and its evaluation with 3D FSCs
• Combatting anisotropy using tilted data collection strategies
• The (unexpected) effect of sampling (in)homogeneity on
global resolution
• Evaluating sampling (in)homogeneity using the SCF criterion
• Considerations for high-resolution
-
Considerations for optimizing cryo-EM resolution• Microscope (e.g. JEOL, TFS, Krios G3/G4, Arctica, etc) microscope settings (e.g. KV, aperture
size), microscope alignment (coma-free, parallel illumination)• Ancillary hardware: cold or warm FEG, energy filter, aberration correctors, phase plates• Detector and detector DQE: Falcon4, Falcon3, K3, K2, CCD, etc• Detector settings: counting vs. super-resolution, dose rate, frame rate• Imaging and data collection: magnification (and pixel size), defocus range, image shift or stage
position, speed vs accuracy (without compromising quality), hole selection, with or w/o stage tilt
• Grid / data collection quality: thick/thin ice, how much drift, sample dispersity, substrate support,grid type
• Image processing:◦ Software (Relion, CryoSparc, CisTEM, Warp/M, EMAN, etc)◦ Initial motion correction and CTF estimation (global, patch-based, per-particle)◦ Classification strategy◦ Post-processing (CTF refinement, particle polishing [movement, B-factor weighting], Ewald
correction, density modification)
• Fourier space sampling• Sample !!!!!!!!!!!!!!!!!!!!!!!!!
-
Nakane, T., …, Scheres, S. (2020). Single-particle cryo-EM at atomic resolution.bioRxiv, 2020.05.22.110189.
Considerations for optimizing cryo-EM resolution:1.2 Å apoferritin (cold FEG)
-
Yip, K. M., Fischer, N., Paknia, E., Chari, A., & Stark, H. (2020). Breaking the next Cryo-EMresolution barrier – Atomic resolution determination of proteins! bioRxiv, 645,
2020.05.21.106740.
Considerations for optimizing cryo-EM resolution:1.2 Å apoferritin (warm FEG)
• Titan Krios (Schottky FEG)• Voltage: 300• Monochromator + Cs corrector• Falcon 3 detector• 0.492 Å/pix• Electron dose (e-/Å^2): 50• Software: Relion• Number of particles: 1,090,676• Resolution: 1.25 Å
-
Considerations for optimizing cryo-EM resolution:1.3 Å apoferritin (cold FEG)
Tegunov, D., Xue, L., Dienemann, C., Cramer, P., & Mahamid, J. (2020). Multi-particle cryo-EM refinement with Mvisualizes ribosome-antibiotic complex at 3.7 Å inside cells. bioRxiv, 44, 2020.06.05.136341.
• cryoARM300 (JEOL)• Voltage: 300• Cold FEG• K2 detector• In-column omega filter• 0.49 Å/pix• Electron dose (e-/Å^2): 88• Software: M• Number of particles: 120,295• Resolution: 1.34 Å
Dimitry TegunovReprocessed dataset in M
EMPIAR-10248
Keiichi NambaData collection (Kato et al, MIcroscopy
and Microanalysis, 2019)
-
Rado DanevYanagisawa-san, Masahide
Kikkawa
Considerations for optimizing cryo-EM resolution:1.3 Å apoferritin (warm FEG)
-
Try to understand your limitationsFirst 2 e-/Å2
Tan, Y. Z., Aiyer, S., et al. (2018). Nature Communications, 9(1), 3628–11.
-
Exposure Dependency of the Mean Intensityof the Unmerged Integrated Reflections
Hattne, J., Shi, D., et al. (2018). Analysis of Global and Site-Specific Radiation Damage inCryo-EM, 26(5), 759–766.e4.
-
• Resolution anisotropy originates from preferred orientation and
inhomogeneity in Fourier sampling
• 3D FSCs provide quantitative measurements of resolution
anisotropy
• Tilts help to overcome resolution anisotropy
• Inhomogeneity in Fourier sampling also causes attenuation of
global resolution: incomplete sampling worse than
inhomogeneous (but complete) sampling
• The Sampling Compensation Factor (SCF) estimates the
amount of global resolution attenuation
• Many factors to optimize for high-resolution … toward sub-Å
reconstructions!
Summary
-
Acknowledgements and Support
Computationalsupport
Phil BaldwinJC Ducom
Funding
Collaborators
Niko GrigorieffBen Himes
Mavis Agbanjie-McKenna
Robert McKenna
Cryo-EM supportBill Anderson
Scripps/SalkConsortium
Jamie WilliamsonTravis BerggrenMartin Hetzer
Leona M. and Harry B. HelmsleyCharitable Trust Grant
Phil BaldwinYong Zi Tan Sriram Aiyer