Job: Patch CTF Estimation
Patch-based CTF estimation.
At a Glance
Estimate the Contrast Transfer Function for micrographs, accounting for the fact that the sample may not be perfectly flat.
Description
Patch-based CTF estimation automatically estimates a defocus landscape for tilted, bent, deformed samples and is accurate for all particle sizes and types including flexible and membrane proteins. This is accomplished by a fast GPU implementation that usually takes 1 - 2 seconds per micrograph. No prior knowledge about particle locations is needed.
Inputs
Micrographs
Patch CTF Estimation requires motion-corrected micrographs, typically straight from a Patch Motion Correction job.
Commonly Adjusted Parameters
The default parameters for this job are often sufficient for high-quality results.
Outputs
Micrographs processed
Micrographs with CTF estimates. During particle extraction, these estimates are used to calculate a local CTF at a particle’s precise location.
Micrographs incomplete
Micrographs which were not successfully corrected are output in this group. Typically this is due to some error specific to the micrograph file — the Event Log should contain more information on why a particular micrograph failed.
Diagnostic plots
Several diagnostic plots are also produced for each of a number of micrographs.
1D search plot
This plot indicates the quality of fit for a 1D search over defocus. Higher values indicate a better fit between the idealized CTF at that defocus and the data. Ideally, a single sharp peak is observed, indicating that there is one defocus that matches much better than the others.
2D defocus landscape
The defocus landscape displays the modeled defocus over the entire micrograph. Typically, this landscape should display some variation that captures the shape of the ice layer.
CTF Fit
The CTF fit helps you assess the quality of your data, and the quality of the CTF fit to that data. There is a lot of information in this plot, so we will build it up piece by piece.
First, the black line is the radial average of the power spectrum of the image. The Thon rings (used to fit the CTF) are present as oscillations between 0 and 1 in this image. This plot accounts for astigmatism using equi-phase averaging (e.g., similar to GCTF [1]).
The red line is the ideal CTF which has been found by Patch CTF Estimation. Note that this is the ideal CTF at the average defocus across the micrograph — each patch will have a slightly different function fitted to it. Ideally, this red line perfectly coincides with the black line.
We can measure how well the power spectrum and the ideal CTF match by calculating the correlation between the two. This is plotted in blue. The point at which this line crosses 0.3 is typically called the CTF fit resolution, and is marked in the plot with a green vertical line. Note that this is not necessarily a hard limit on the quality of your data, and the images are not filtered to this resolution — it simply gives you an estimate of what quality to expect from this image.
The final image, with all three of these lines, is displayed below. Note that the plot title also contains some useful information
The defoci on the major and minor axes are reported in Å (DF1 and DF2, respectively)
The astigmatism angle, in radians, is given (ANGAST)
The phase shift (if any) is reported (PHASE)
The fit resolution is reported in Å (FIT)
Ice thickness
The relative ice thickness is measured by comparing the background signal in a band centered on 0.265 Å-1 (indicated by a blue fill) to a wider band which includes this region (indicated by the green bars). If this band has high background, it means there is more ice in the image which, assuming ice takes up approximately the same proportion of each image, means the ice is thicker.
This plot does not directly report anything about data quality or CTF fit quality, although it is generally taken to be true that thinner ice results in higher quality reconstructions.
Common Next Steps
Micrographs with CTF estimates are ready for particle extraction — a common workflow is performing Blob Picking and Particle Extraction at first, then using the initial results for more advanced particle picking jobs.
After a high-quality map is achieved, data may benefit from CTF Refinement to improve defocus estimates and account for higher-order aberrations.
Implementation Details
The overall implementation of Patch CTF follows a similar pathway to Patch Motion Correction. First, a coarse estimate of the CTF is modeled for the entire micrograph. Then, patches are used to estimate a function which returns the modeled defocus for a given micrograph coordinate.
For an initial, coarse estimate of the CTF, it is assumed that there is no astigmatism. A simple correlation with the radially-averaged power spectrum is used to find the best-fitting defocus.
A new envelope function is then calculated using this coarse defocus estimate, after which the 2D CTF is estimated for the entire micrograph including astigmatism. The estimated defocus is refined for each patch.
These patch CTF estimates are used to fit a spline function which provides the estimated defocus at a given (x, y) coordinate on the micrograph. Ultimately, particle defocus estimates come from this spline function. The Override knots y and Override knots X parameters control the degrees of freedom of this spline function. Note that changing the number of knots does not change the number of patches used by Patch CTF Estimation. See the Patch Motion Correction job for a more detailed explanation of these parameters.
References
Zhang, K. Gctf: Real-time CTF determination and correction. J. Struct. Biol. 193, 1–12 (2016).
Tegunov, D. & Cramer, P. Real-time cryo-EM data pre-processing with Warp. bioRxiv (2018) doi:10.1038/s41592-019-0580-y.
Last updated
