Nimbus downstream analysis

[joaolsf]

Description of feature

Hi, I wanted to ask this directly to the developers of this package and I could not find a nimbus specific topic in the scverse discussion community.

I am having the first cell tables extracted from Nimbus and I tried to find this info in the paper and here but might not be very clear. Just wanted to ask you:

1- The cell table does not output the cell area, the x and y coordinates. The companion ark-analysis method, developed by the same group, exports cell physical features such as size and coordinates. Are the "label" columns exported in the nimbus cell table and in the ark-analysis cell table, matched? If so, do I have to run both methods to extract the physical aspects of each cell, including their x, y coordinates?

2- How do you suggest to use the marker confidence score values downstream? Use as is, use as z-scores (like in the paper)? I assume there is no “normalisation” that should be done like we do with protein expression matrixes generated with other methods. My understanding is that Nimbus does not output the “actual" protein expression, these are the confidence score of expression, correct?

3- My data has isotype controls, such as mouse and rabbit IgGs. Do you recommend subtracting the confidence scores for these controls antibodies out of the panel protein markers prior to any downstream analysis?

If there is a channel specific for nimbus in the scverse, please direct me there.

Thanks!

[JLrumberger]

Hi @joaolsf,

thanks for the detailed questions. Here are my answers:

1. As far as I know, the label columns are matched in Nimbus and Arc and reflect the ID of the cell segment in the provided cell segmentation masks. What's your opinion on the x,y coordinates and the cell size, should they also be included in the Nimbus output file because they'll be needed downstream anyways? I'd be happy to include them if you think this improves usability.

2. We used z-score normalization for visualization, and it is also used for the GUI in the PIXIE cell clustering pipeline. Exactly, Nimbus returns confidence scores for marker positivity. Therefore, for downstream analysis, i.e. the SOM that is trained within the PIXIE pipeline for cell clustering, we used unnormalized nimbus scores.

3. No, I assume these isotype controls should control for non-specific binding (?), then I think these should be subtracted from the input data prior to feeding it into nimbus in order for background correction.

All the best,
Lorenz

[joaolsf]

Hi @JLrumberger ,

Thanks for the quick answer and details.

1- I will use ark-analysis to extract the "integrated expression" anyway but would be good to have the option to extract the cell size x, y coords and any other physical cell parameter in nimbus in case the user chooses to only use the confidence scores from nimbus.

2- Understood!

3- Yes, I agree with you! However, there is always this discussion, and some people are from the "background correction school". I've discussed with the comp bio team of the company from where I am getting the data from and they confirmed they do not recommend background subtraction using the isotype controls.

Thanks,
Joao

[JLrumberger]

Hi @joaolsf,

thanks for the answer. Then I'll add the x,y coords and cell size in pixels to the nimbus readout and close this issue afterwards.

All the best,
Lorenz