RNAScope Quantification for Beginners

Hi fellow scientists,

I am a master student (trying to graduate) and have had good success with RNAScope. However my PI wants to quantify the respective images but I’ve never done it before.
Therefore I would very appreciative for a step-by-step and opinions of which programs would be best for counting individual transcripts and dual/triple labeled cells. Also which graphical representation would be best for relating this information to scRNA-seq results?

Thank you in advance and I look forward to great suggestions.

Is your RNAscope analysis in brightfield (red/teal?) or fluorescent images?

There are several other posts here about quantifying that type of staining, and the problems involved in it (mostly on the brightfield side).

Mine are fluorescent images.

Excellent start! Unfortunately, I can’t really help with the CellProfiler part, as I generally use QuPath. It’s pretty good for tissue analysis with constraints on cell expansion, but not so good for large, sprawling, or oddly shaped cells. YMMV.

Are you able to pair the scRNA seq with exact cells in your images (laser microdissection)? Or are you looking at relative populations from isolated cells and tissue images?

1 Like

I’ve never heard of QuPath, but I will give it a try. Thank you very much!
Also to answer your question I’m looking at specific populations from tissue images. What kind of data can one extract using the QuPath approach? (mean number of transcripts per cell, number of dual/triple labeled cells)
What is a significant amount of labeled cells normally used in analysis?

Yowch, that might be a bit high (of an ideal target!). Ideally you have a spot count, and if you are very lucky or looking for a very rare transcript, it might be accurate, that single spots correlate with single transcripts. More often it is somewhat of a mess, and you are looking at relative rates, given similar errors between samples.

QuPath can give you a rough measure of the rate of stain appearing within the area around a particular nucleus, ideally determined to be a “cell.” This is never really accurate, though you could only do better if you have some kind of membrane marker and a more advanced algorithm where you could actually determine the borders of a cell. In the end, you can end up with a population that have a rough # of spot counts per cell.

I would recommend a “low” “medium” and “high” amount of spots, determined by your actual spot population. And then compare from there… but it would be somewhat experiment dependent.

If I understand correctly, your goal is to identify and count dots associated with each cell, in a tissue? If so, the place to start with CellProfiler is our example pipeline for exactly that:

I suspect the key things you will need to adjust are:

  1. Identifying the nuclei properly. You will want to adjust the settings in the IdentifyPrimaryObjects module until it is accurate. There are lots of tutorials to help you get the hang of this, or you can post an example image and the pipeline you’ve attempted here to get help adjusting.

  2. Associating spots to cells. The example pipeline expects all dots to be in the nucleus, so if you want to associate dots that are outside nuclei, just add an IdentifySecondaryObjects module right after IdentifyPrimaryObjects. You can use it to define the cell borders, then you can Relate dots to cells instead of dots to nuclei. For more about understanding primary objects (nuclei) and secondary objects (cells), see our short blog post (ignore the security warning).

Let us know if you run into trouble!


Hi Anne,

Yes that is correct, I am trying to identify labeled cells as well as count the individual dots/transcripts.
Thank you so much for the detailed instructions. I am currently trying the speckle pipeline (along with adjusting the segmentation) and I’ll let you know the outcome!

1 Like

Thank you Research_Associate for taking the time to share your image analysis approach!