RNA detection with FISH-quant

Two options exist, you can either directly analyse the smFISH channels in FQ to detect individual RNAs, or you can use a Matlab script to detect individual RNA molecule and then decomposes RNA aggregates into individual RNAs with a Gaussian Mixture Model (GMM).

Detection of individual RNAs

Below only a quick summary of the workflow is presented. For more details consult the FQ manual.

  1. Open FQ
  2. Set analysis folder to folder containing the image that should be analysed: Menu Folder > Set root folder)
  3. Open image that should be analysed.
  4. Draw outline of embryo / cells:
    1. Open dedicated interface: Button Define outlines
    2. Draw a new cell: Button New cell
    3. Save outline: Button Quick-save
    4. Return to main FQ: Button Finished
  5. Filter image (default filter with LoG is usually good).
  6. Inspect image. On the right part of the interface select Filtered image, disable outline, and double click on the image. This will show a maximum intensity projection of the filtered image in a separate window. Here, you can change the contrast and zoom. In order to determine an appropriate threshold in the next step, determine a generous range of intensities range corresponding to the individual RNA molecules.

  7. Set pre-detection settings. FQ will test how many RNAs are detected with a range of user-defined intensity thresholds (based on the manual inspection of the image from above).

    1. Open dedicated interface: Button Detect
    2. In the user-dialog change the first two parameter: minimum and maximum threshold to test. We usually use a minimum value that’s somewhat lower than the lowest RNAs to also consider intensity value corresponding to background.
    3. FQ will calculate the number of detected RNAs for a range of values in this interval. Depending on the size of the image, this can take a little while. Once done, a plot with the number of detected RNAs as a function of the different thresholds is shown. If the specified range is not appropriate, it can be changed in this interface the computation be repeated.
    4. Ideally, this curve shows a plateau for a value range corresponding to a good detection threshold. If that’s not the case, a reasonable value corresponding to a visual assessment of what individual RNAs are should be chosen. A pre-detection with a given detection threshold can be perform and the results be inspected in a separate window.

    One other parameter that could be important to adjust is the cropping region around each RNA that is considered for analysis and fitting. Reducing this size allows sometimes to better detect closely spaced RNAs but at the cost of an imprecision in the fit. For most applications a value of +/- 2 voxels in XY and Z are a good compromise.

  8. Once you are satisfied with the settings press on Perform detection for all cells to return the main window.

  9. Press button Fit to fit each detected spot with a 3D Gaussian function.
  10. [Optional]. You can set threshold values for the different fitting parameters, e.g. remove spots with very small or very large standard deviations.
  11. You can now save the detection settings: [FQ] Main>Save>Save detection settings. As a file-name specify FQ__settings_MATURE.txt
  12. Depending on which analysis you want to perform, you either save the detection results or only the specified cell outline.
  13. If you are satisfied with these detection results, you can save them directly: [FQ] Main>Save>Detection spots [All].
  14. If you would like to also analyse RNA aggregates, you can save only the outline: [FQ] Main>Save>Outline of cells.

Detection of clustered RNAs

Her we use an approach that detects individual RNAs and then decomposes aggregates into individual RNAs (for more information see our Nature Communications paper). This requires that the detection of individual RNA as described above has been performed.

After performing the steps above you will have two text files (1) the cell outline, (2) FQ detection settings. The latter can be for an entire experiment or for individual outline files. Depending on how your data is organised a different workflow below will be adequate

Detection settings for individual outline files

Here, each outline has it's own settings file. We provide a script on the FQ BitBucket repository to analyse these data (WRAPPER_smFISH__GMM_v1.m).

  1. The scripts requires that images, outlines, and settings are in the same folder.
  2. You have to specify a string that identifies outline files, and a string that identifies the settings files. Importantly, replacing one string by the other should then yield the respective other files.
  3. The script will then loop over all outlines, and process them with the GMM approach
  4. Results will stored in a newly created folder with the same name as the outline file. Here two subfolders are created,
  5. results_GMM contains the RNA detection with the GMM (to decompose foci into individual RNAs). The most important files are the text files ending with _res_GMM.txt containing the RNA detection results, which will be used to perform the co-localisation. The folder further contains a subfolder #plots with a PDF showing the results of the GMM detection. Individual RNAs are shown in green, decomposed foci with a red number indicating how many RNAs were estimated to be in each foci.
  6. results_noGMM contains the RNA detection results without the GMM.

In the example below, the strings would be outlines.txt for outlines, and settings_MATURE.txt for the settings. It also shows the newly created folder, containing the GMM results.

.
├─ fish_data/
│  ├─ exp1_ch1_pos01.tif                  # smFISH image
│  ├─ exp1_ch1_pos01__outlines.txt         # FQ outlines results
│  ├─ exp1_ch1_pos01__settings_MATURE.txt  # FQ detection settings
│  ├─ exp1_ch1_pos01__spots.txt            # FQ spot detection
│  ├─ exp1_ch1_pos01__outlines/
│     ├─ exp1_ch1_pos01__outlines/
│        ├─ results_GMM
│           ├─ exp1_ch1_pos01_res_GMM.txt
│           ├─ ...
.