HPA Cilia Study - Code repository

This repository contains instructions and programming code for analyses of the data presented in the preprint "Intrinsic Heterogeneity of Primary Cilia Revealed Through Spatial Proteomics (DOI: 10.1101/2024.10.20.619273)". Please cite our preprint if you reuse data or code.

We also provide a license in this repository for all code where the license is not otherwise specified in the code.

If you do not see this folder system on github, check the following github repo for potential updates: https://github.com/CellProfiling/HPA_Cilia_Study_Code/tree/main.

(1) Downloading data from the protein atlas

1. Query the XML data for the ENSEMBL gene ID of your interest through a link composed of https://www.proteinatlas.org/ + ENSEMBL ID (e.g., ENSG00000137691) + .xml, e.g., https://www.proteinatlas.org/ENSG00000137691.xml for CFAP300.
2. In the XML search for the subassay with subtype ("ciliated cell lines"): <subAssay type="human" subtype="ciliated cell lines"> ... </subAssay>

3. Nested in the subAssay element you will find different "data" elements, which each represent the images for a specific cell line.

4. This allows you to see the links for all z planes. E.g., for the image in the screenshot, this reveals 21 different z planes with their corresponding links:

   1. For z_index=1: https://images.proteinatlas.org/38585/2146_D7_42_blue_red_green.jpg
   2. For z_index=2: https://images.proteinatlas.org/38585/2146_D7_41_blue_red_green.jpg
   3. ...
   4. For z_index=20: https://images.proteinatlas.org/38585/2146_D7_24_blue_red_green.jpg
   5. For z_index=21: https://images.proteinatlas.org/38585/2146_D7_23_blue_red_green.jpg

5. Extract a download id for each z plane from the link:

   1. Remove the beginning (https://images.proteinatlas.org/) from the link. E.g., https://images.proteinatlas.org/38585/2146_D7_42_blue_red_green.jpg becomes /38585/2146_D7_42_blue_red_green.jpg
   2. Remove the ending ( blue_red_green.jpg) from the remaining link. E.g., /38585/2146_D7_42_blue_red_green.jpg becomes /38585/2146_D7_42_
   3. The final image id (/38585/2146_D7_42_) remains

6. Recombine image id for each z plane to create links for downloading the individual tif images for each channel.

   1. For the blue DAPI / Nuclei channel, the download link will be: https://www.proteinatlas.org/download_file.php?filename= + the image id (e.g., /38585/2146_D7_42_) + _blue&format=tif.gz, so e.g., https://www.proteinatlas.org/download_file.php?filename=/38585/2146_D7_42_blue&format=tif.gz
   2. For the red Cilia marker channel, the download link will be: https://www.proteinatlas.org/download_file.php?filename= + the image id (e.g., /38585/2146_D7_42_) + _red&format=tif.gz, so e.g., https://www.proteinatlas.org/download_file.php?filename=/38585/2146_D7_42_red&format=tif.gz
   3. For the yellow Basal body marker channel, the download link will be: https://www.proteinatlas.org/download_file.php?filename= + the image id (e.g., /38585/2146_D7_42_) + _yellow&format=tif.gz, so e.g., https://www.proteinatlas.org/download_file.php?filename=/38585/2146_D7_42_yellow&format=tif.gz
   4. For the green Protein of interest marker channel, the download link will be: https://www.proteinatlas.org/download_file.php?filename= + the image id (e.g., /38585/2146_D7_42_) + _green&format=tif.gz, so e.g., https://www.proteinatlas.org/download_file.php?filename=/38585/2146_D7_42_green&format=tif.gz

7. Download for all z planes all four channel .tif.gz files and place them all together into one folder.

8. Extract all .tif.gz files so they become .tif files and rename each file based on the z plane and the channel to match the following scheme: <image id after removing the front part between / and /> + _c + + _z + <z_index> + .tif, such as, e.g., 2146_D7_42_c0_z0.tif for the DAPI channel image of the first plane for the example images above.

   1. Use channel number 0 for the DAPI channel
   2. Use channel number 1 for the Cilia channel
   3. Use channel number 3 for the Protein of interest channel
   4. Use channel number 4 for the Basal body channel

9. Assemble all images into a multi-channel multi-plane tif stack.

(2) Segmenting the images

Images downloaded and assembled as described above can be subjected to the segmentation script to create cilia, basal body, and nucleus segmentations.

(3) Run CiliaQ analysis on the segmented images

1. Run CiliaQ analysis (CiliaQ version V0.2.1, which needs to be manually installed (not through ImageJ update sites)) on all 7-channel images with the following CiliaQ settings

2. Collect all CiliaQ output files ending with _CQs.txt in a folder.
3. Add the table legend file provided here to the folder with all CiliaQ's _CQs.txt output files. Make sure that this legend file is the first item when alphabetically sorting all files in the folder (if not rename to have it become the first item while keeping the file ending _CQs.txt intact)
4. Finally, assemble all _CQs.txt-ending files produced by CiliaQ through concatenating all files ending with _CQs.txt into a single text file, e.g. by adding this Windows batch file to the folder and executing it; it will produce a new file called AllCQsFilesMerged.txt with all concatenated CQs.txt files, which can then be used for analyzing statistics on the cilia like the length or orientation angle (as shown in Figure 2 in the preprint).

(4) Extract and normalize intensity profiles

1. Use the AllCQsFilesMerged.txt file from the previous step for this analysis.
2. Collect all _CQl.txt files created by CiliaQ in the previous steps in a folder.
3. Run the analysis as described in the readme file and use the scripts in this repository
   1. Note that you need to have a specific excel file for this that lists all the images that you want to include and has specific columns available as explained in the readme file.
   2. The output files from this analysis can be further used to cluster and analyze intensity profiles.
4. To plot intensity profiles follow this readme file and use the scripts in this repository
5. To cluster profiles follow this notebook

(5) Training cell cycle prediction model and applying it to images

• See readme file with instructions in this subfolder.
• A pretrained model file is available here: .

(6) Analyze cell cycle prediction data

1. Validate GMNN prediction values based on predicting on images with other cell cycle markers stained in an additional channel (related to the preprint Figure 4H). A jupyter notebook showing how to normalize intensities and plot the predicted nuclear GMNN intensity versus the real intensity in the protein channel as well as running statistics is shown here
2. Combine measures of nuclear predicted GMNN intensity with CiliaQ measurements of ciliary intensity (or other cilia parameters). The code is producing the preprint Figure 4I). A jupyter notebook showing how to normalize intensities and plot the predicted nuclear GMNN intensity versus the real intensity in the protein channel as well as running statistics is shown here; in case you favor using a python script you can find the same code as in the jupyter notebook here.

(7) Other bioinformatic analysis of protein lists

Note: Html files provided in this section represent knitted R markdown files. Download them and open in browser to explore them.

• Figures 1I and 2E (and related supplemental tables)
  • Creating a database about cilium-associated genes from existing data sets and studies
  • Comparing the cilia atlas and its annotations to the database (creating Figures 1I and 2E and related supplementary tables)
• Figures 3A, 4C, 4D (and related supplemental figures / tables): Functional enrichment analysis of protein lists
• Figure S1C: Test if the distribution of the number of subcellular locations of ciliary proteins is significantly different from all proteins in the whole cell