MCMICRO compatibility with in situ sequencing

[rocketeer1998]

Hi @jmuhlich, thanks to you and your team for providing this easy-to-use software to the community! I'd like to try MCMICRO on my ISS (in situ sequencing) data, which belongs to a branch in spatial transcriptomics.

What I have done

• [x] Successfully install and configure MCMICRO on my server
• [x] Successfully test run on exemplar-001 data, and the output folder structure is like:

The data structure of ISS

Briefly, ISS detects mRNA on the whole tissue slides using cyclic imaging and barcode decoding.

Input1: raw image files (in .tif format)

The raw images obtained from an ISS experiment are almost the same as that of exemplar-001, but only differ in the:

• ISS images contain fluorescent dots, each representing a unique mRNA in essence. I wonder if this characteristic would lower the accuracy of stitching, registration, and all other downstream analysis.
• Image files are in .tif format, rather than .ome.tif.

Input2: markers

This is the screenshot of my markers.csv.

Question

I've tested MCMICRO on my ISS data, however, it failed on quantification:mcquant step which alerted, Exception: The number of channels in markers.csv doesn't match the image. Stitching and alignment can run quickly, despite the output was incorrect after visual inspection.

Thanks for your patience!

[rocketeer1998]

Maybe I should specify the --align-channel parameter in ashlar to 5 because I want to align based on DAPI channel. Is that right?

[jmuhlich]

How many input tiff files do you have and what are the image dimensions? Do you have one whole-slide image for each mRNA transcript, or do you have many smaller "tiles" for each transcript?

[rocketeer1998]

• There are 50 FOV (fileds of view) and each has 5 tiff files according to 5 channels (Cy5, Cy3, TXR, Cy7, DAPI), with [2048,2048,3] dimensions. All images are obtained using Leica fluorescence microscope and respective LAS X software.
• I have many smaller 'tiles' for each transcript and I'd like to perform stitching, registration, and spot detection among them.

[Boehmin]

I wanted to follow up on this thread since I was wondering whether mcmicro will support a module for ISS decoding across multiple rounds, or whether that is still in the pipeline? Thank you!

[ArtemSokolov]

What software do folks currently use for ISS decoding? If there is an existing containerized module we can leverage, it would be relatively straightforward to add one to the pipeline.

[Boehmin]

I guess spacetx/starfish offers a lot of flexibility (in terms of different decoding schemes) - https://spacetx-starfish.readthedocs.io/en/latest/user_guide/index.html#spot-based-decoding.

Additionally, I must admit I have not been able to figure out how to implement spot detection with 3seg yet - I`ve been trying a few params.yml modifications I found https://mcmicro.org/troubleshooting/tuning/s3seg.html, however, I keep getting an error that the puncta parameters are not supported. Is there an optimum configuration for the params.yml required for spot detection to work?

Thanks ever so much for your help!

[clarenceyapp]

Hi @Boehmin Can you list the parameters you are using for s3seg spot detection please? Thanks.

[Boehmin]

I added the following to my s3seg options: s3seg: --maxima-footprint-size 15 --area-max 50000 --expand-size 9 --pixelSize 0.18 --mean-intensity-min 50 --detectPuncta 2 3 4 --punctaSigma 1.5 2 1.75 --punctaSD 3

[clarenceyapp]

Hi @Boehmin , thanks for listing your s3seg parameters. Your spot detection parameters look valid. However, it appears you're using the newer version of s3seg and unfortunately, we have not had a chance to port the spot detection capabilities that were in the older version yet, which is why it is erroring. What are the image dimensions in terms of number of pixels in X and Y? If they are small enough (~30,000 pixels), the older version will work fine with a few additional changes to your other s3seg parameters.

[Boehmin]

Hi @clarenceyapp , thanks a mill for your help - that`s great news! I use s3seg version: 1.5.4-large if that helps as well. Which version do you recommend I go for?

The image is 21,869x17,528 pixels so should be ok then?

[ArtemSokolov]

Another thing to note is that s3seg would only detect the x-y coordinates, so additional processing would need to happen for trace building and spot detection.

It seems that spacetx/starfish is a good one-stop-shop for it. If anybody is able to wrap something like this into a command-line tool that works with .ome.tiff, I would be happy to containerize it and include it in MCMICRO. Anything with an interface like

python starfish.py --in myimage.ome.tiff --codebook codebook.csv

would work.