Postprocessing

Once the contact maps have been computed, you often need quantification and additional filtering. For example, coverage, features descriptors, and so forth.

There are three key processing steps disjoint from the actual algorithm output:

• bleedthrough filter
• CSV curation
• Sampling

In the remaining of this document, let us assume DIR is the directory where the algorithm saved its output on a full dataset.

Bleedthrough filter

The background filter removes ghost effects (bleedthrough). It is run as part of the pipeline, but you can invoke it separately.

Suppose we want to filter all tif files ending with "1.tif" or "2.tif" , for z=1 to 1.1 in 0.25 steps, and then compute the object properties.

julia --project=.  scripts/segment.jl --inpath mydir -z 1.0 -Z 1.1 -s 0.25 -r "*[1,2].tif"

For each file, for each filtered value, it will generate:

• mask.tif
• masked.tif

For all the files, it will generate a CSV with columns, where each row is an object:

• size (nr of non zero voxels)
• weighted (intensity sum of objects)
• minimum, Q1, mean, mediam, Q3, maximum, std, kurtosis : describes the intensity distribution of the object
• xyspan : major axis of 2D projection, pixels
• zrange : extent in Z
• zmidpoint : midpoint in Z slice
• distance_to_centroid: distance of this object's centroid to centroid of all objects, describes clustering
• distance_to_centroid_normalized: rescale the above to 0-1, where 0 is centroid of all objects, 1 is maximum dispersion
• centroid_channel_{x,y,z} : in pixel coordinates, records the centroid of all objects, this is the reference for the distance computation
• centroid_object_{x,y,z} : in pixel coordinates, records the object centroid
• filename : the source tif file name
• z : the z value used
• eig1-3: PCA eigenvalues, the can be used for shape descriptors.
• eig1-3normalized: eigenvalues rescaled to 1.

CSV Curation

You can run our Python scripts to aggregate and curate the processed CSV files.

python3 scripts/csvcuration.py --inputdirectory <where you saved the output> --outputdirectory <where you want the new CSV files saved>

By default this will look for output produced with $\alpha$ 0.05, you can override this as needed with --alpha 0.01 for example.

Parameters

• –lnsize: vesicle filter size for mitochondria, default set to 9, override to zero if you want no filtering
• –mitoint: vesicle filter intensity for mitochondria, default set 0.2, override to zero if you want no filtering

This will produce:

contacts_aggregated.csv             # Contacts aggregated per cell, so 1 row = 1 cell, use this for e.g. mean height, Q95 Volume
contacts_filtered_novesicles.csv    # All contacts, without vesicles
contacts_unfiltered.csv             # All contacts, no filtering

Sampling contacts

In scripts/run_cube_sampling_on_dataset.jl you'll find a script that samples contacts with a sliding window, to avoid long tail statistics dominating the conclusion of any analysis. The paper goes into more depth why this is beneficial.

julia --project=. scripts/run_cube_sampling_on_dataset.jl  --inpath DIR --outpath  <where to save your output>

This script takes extra parameters as needed:

• "–cube-vesicle-size-ln": The size of objects to ignore (anything <= will be ignored in the first channel). Default 9. Set to 0 if first channel is not mitochondria.
• "–cube-vesicle-sample-size": The sliding window size in voxels (cube), default 5, used to sample the data.
• "–cube-vesicle-intensity-mean" : The intensity of objects that should be ignored in the first channel (< than this will be dropped. Default 0.2.) Set to zero if the first channel is not mitochondria

A convenience script is provided to further aggregate the output of this stage.

python3 scripts/coverage.py  --inputdirectory DIR --outputdirectory <where to save your ouput>

You may need to switch python3 with python, depending on your installation.

This will print summary output and save a file coverage_aggregated.csv. The columns Coverage % mito by contacts, mean per cell and ncontacts mean are the columns you'll be most interested in.

They report the coverage of contacts on mitochondria (minus MDVs), and the number of contacts per sliding window of 5x5x5 voxels.