How to calculate nuclear:cytoplasm ratio for a protein (RhoE) for H-DAB images

Hi there,

I’ve gone through the tutorial on the QuPath website and on youtube. What would be the best way to analyse the expression of an intracellular protein (RhoE) in the nucleus and cytoplasm stained using H-DAB staining?

Since it is DAB and not fluorescence, the primary analysis method is area since intensity cannot be measured linearly. You could try an independent 1+ 2+ 3+ method for nucleus and cytoplasm, but that would be rather strongly influenced by Qupath’s blind cytoplasmic expansion. Finally, you haven’t posted an image, but is QuPath able to do a good job segmenting the nucleus with DAB staining obscuring both the nucleus and cytoplasm?

This sounds like a better experiment for IF, but maybe you could give more details and an example image.

See here for the limitations of DAB staining.

  • Some histological dyes scatter light considerably, rather than absorbing it (e.g. silver impregnation methods) and therefore cannot be unmixed reliably.
  • Staining intensity quantification is meaningful only when dye uptake is stoichiometric (e.g. Feulgen reaction for DNA). Immunohistochemical methods are unfortunately non-stoichiometric (like most histological dyes), making them unsuitable to measure ‘antigen expression’ based on DAB intensity. The main reasons are:
  • Antigen-antibody reactions are non-stoichiometric, therefore “darkness of stain” does not equate to “amount of reaction products” or “expression” of a given antigen. In fact most histological stains are non-stoichiometric (one of the known exceptions is Feulgen stain, commonly used for DNA cytometry).
  • Immunohistochemistry also uses a series of amplification steps to visualise the results, making it difficult to control what the final intensity of the amplified signal actually represents in terms of amount of antigen.
  • The chromogen DAB does not follow Beer-Lambert law. See, e.g., the paper by CM van der Loos:
    “The brown DAB reaction product is not a true absorber of light, but a scatterer of light, and has a very broad, featureless spectrum. This means that DAB does not follow the Beer-Lambert law, which describes the linear relationship between the concentration of a compound and its absorbance, or optical density. As a consequence, darkly stained DAB has a different spectral shape than lightly stained DAB.”
    Further concerns have been raised in this message posted by Al Floyd to the ImageJ mailing list.

Other image analysis programs might be better if you can find one that reliably traces the outlines of your cells - though again, you cannot use mean OD to compare the nuclear and cytoplasmic expressions directly.

1 Like

Hi there,

That is very helpful! I don’t suppose you would be able to have a short MT meeting to explain QuPath in a bit more depth? I’m a novice to this and haven’t really got anyone who knows better. I’ve been through Pete’s tutorials but am still confused as to some of the functions and how/when to use them

Let me know if you are able to schedule a meeting/any other recommended tutorials?

Thank you

1 Like

Sorry, I suspect your best course of action is to create a detailed post of what you want to accomplish and post it here. Pictures showing what you have, and what you want to accomplish, along with the final values and very generally why are all useful. The list included in any initial post is a good start.

No worries, I should have done that originally, apologies!

Essentially, what my team and I are trying to do is quantify the number of fibroblasts per square micron from skin sections in the normal vs diseased states. We then also want to quantify the expression of a protein (RhoE) that is found in fibroblasts and compare the expression levels in fibroblasts to see if there is a significant difference in RhoE expression in the diseased state. We are also trying to quantify the expression levels of 2 other cytosolic proteins - NICD and PLOD2. I have attached images of the skin samples. I shall be using the annotation function to segregate the epidermis and dermis (you can see the border between them in the image)

Please let me know what you think is the best course of action.

Much appreciated

The image helps, though exporting a region as an OME-TIFF at full resolution allows people to provide more useful feedback. Even with that png, I was able to quickly get:

This is replacing the Positive cell detection threshold that by default is Nuclear with Cytoplasmic. You may instead want whole cell, I am not sure. As @Ajay_Zalavadia pointed out, all cytoplasmic markers are also “nuclear” markers when you clip the top or bottom of the cell with your tissue slice (or at least obscure the nuclear staining). I am not sure how important that is to you, and the frequency can depend on the size of the cells/nuclei.

Another thing to keep in mind is that the “mean” cytoplasmic DAB OD will be influenced by the cytoplasmic expansion. Strongly. So play with your Cell expansion slider to get something that is as biologically relevant as possible - or try subcellular detections.

Aside from that, Pete’s IHC YouTube guide is pretty much all you would need - just change the compartment you are classifying.

To reiterate, you cannot quantify the amount of protein from DAB staining, if you want to do that, you will want something more like IMC or fluorescence - they provide more (although not perfectly) quantitative outputs. DAB is great for % positive or positive per ^mm, not so great for checking elevated expression “levels.” Not impossible, but you are relying on arbitrary thresholds for 1+2+3+ and some sort of count.

“Expression is twice as high in X” is going to be impossible.

1 Like

Thank you! Very helpful!

Well we are analysing the expression of the proteins as a % of cells. What is DAB OD, if you wouldn’t explaining?

Many thanks

Optical Density - Log scale.

Here is a random example from the internet - NanoHybrids - Optical density, absorbance & extinction of gold nanoparticles

1 Like

Ah, so is it just a measure of the intensity of DAB staining?

Intensity is a tricky term to use here, because it can mean the strength of something, which would be accurate from a language point of view, but in microscopy/image analysis it generally means how bright something is. OD is more like the inverse of the intensity, with a log scale thrown in. You are looking at how dark the stain is.

You are very helpful, thank you. Just for my own understanding (limited as that may be at the moment!)…

From what I can understand, to try and quantify the expression levels, or relative expression levels of the proteins between normal and diseased skins samples, the fir step would be to first annotate the regions of interest: in this case the dermis and epidermis. Then create a thresholder on Haematoxylin to identify the nuclei. Then create a DAB thresholder on the dermis selection and run “positive cell detection” and use “measure annotations” under the measure drop-down menu to obtain results?

Apologies for my questions, I am a complete novice with QuPath and it seems that the more videos I watch on it, the more confused I get!

Aside from the annotation creation, I recommend going through the IHC tutorial on YouTube - it has all of the steps you will need. You only need to change out the points mentioned above. There are many steps, and it can be easy to get confused at first, but the information you need is all there.

I will do! Thank you. What institute are you working at? I’m working with the Randall Centre at King’s College London. A pleasure to meet you!

Hi there,

I’ve tried watching the videos but am still unsure and scratching my head. I don’t suppose you could spare 20 minutes or so to have a Microsoft Teams meeting to explain if you have the time. Would be very much appreciated!



Afraid I tend to mostly stay anonymous. I do know that there are plenty of experienced and friendly QuPath users around the UK, and at least a few at King’s College, so hopefully you can find someone to help. If you are having trouble with Pete’s videos, though, I think you might want to consider collaborating or asking someone specialized to help or do the analysis for you.

At least a couple of people have posted requests for general QuPath tutoring, though I am not sure what they came up with.

Good luck and good science.