Skip to content

Topic 9: Particle averaging

  1. You are given a dataset (available via Brightspace) which contains a list of 2D coordinates (localization data). Each point represents a single molecule location in nanometer.
  2. Your first task is to render this list of coordinates to a pixelated 2D image with proper size. You can choose one of the methods that are suggested in section “Visualization” of the paper by B. Rieger et al.. The image should show dotted rings. These rings are Nuclear Pore Complexes. Each of your cells has about 1000 of them in the Nuclear Envelope, separating the nucleus from the cell cytoplasm.
  3. Once you have a single large image with hundreds of NPCs, you need to segment and crop each individual NPC from the large image.
  4. Complete your template matching code in labwork 4 in order to segment as much unique NPCs as you can out of the large image.
  5. Now, design a pipeline to first register all the segmented particles (NPCs) and then average them in order to get a single reconstruction.
  6. This single reconstruction which we call it superparticle should have a better resolution than each individual NPC.
  7. You can assume that the NPCs are only rigidly (rotation and translation) different.

  8. Use two different rigid image registration approaches, in order to perform the registration tasks. Preferably, one intensity based method and one feature based method. You don’t need to implement these methods and you are allowed to use existing code for this part. See here as an example.

    Important

    Remember that you are only allowed to use existing pairwise registration algorithm for this groupwise registration task. Think about the way you can use pairwise registration for registering multiple instances.

  9. Compare and discuss the results of the two approaches with the results in Löschberger et al..

  10. Use Fourier Ring Correlation (FRC) to evaluate the resolution of your final reconstruction. Briefly for FRC calculation, you need to first split your dataset into two halves with the same number of particles. Then register them separately to get two superparticles. Finally align these superparticles and use them as input arguments to the function frc() in the mentioned software package.

References

B. Rieger, R. Nieuwenhuizen and S. Stallinga, “Image Processing and Analysis for Single-Molecule Localization Microscopy: Computation for nanoscale imaging,” in IEEE Signal Processing Magazine, vol. 32, no. 1, pp. 49-57, Jan. 2015, doi: 10.1109/MSP.2014.2354094.

Löschberger A, van de Linde S, Dabauvalle MC, Rieger B, Heilemann M, Krohne G, Sauer M. Super-resolution imaging visualizes the eightfold symmetry of gp210 proteins around the nuclear pore complex and resolves the central channel with nanometer resolution. J Cell Sci. 2012 Feb 1;125(Pt 3):570-5. doi: 10.1242/jcs.098822. PMID: 22389396.

Last update: 2023-04-17