Help needed! Learning to use Fiji-coloc2

Hello guys, I was able to download Fiji and have access to coloc2.

However, I am learning how to use this tool. I am selecting the ROIs and tried to play with background subtraction but I still receive the same outcome.

For my first test I received the following message: y-intercept far from zero
The ratio of the y-intercept of the auto threshold regression line to the mean value of Channel 2 is high. This means the y-intercept is far from zero, implying a significant positive or negative zero offset in the image data intensities. Maybe you should use a ROI. Maybe do a background subtraction in both channels. Make sure you didn’t clip off the low intensities to zero. This might not affect Pearson’s correlation values very much, but might harm other results.
Threshold of ch. 1 too high
Too few pixels are taken into account for above-threshold calculations. The threshold is above the channel’s mean.
Threshold of ch. 2 too high
Too few pixels are taken into account for above-threshold calculations. The threshold is above the channel’s mean.

What Am I supposed to do?

1 Like


Welcome to the Forum!

The quick-and-dirty answer is: Proceed with Caution … Colocalization is a tricky analysis for sure - there are many pitfalls - so study up! The Colocalization Analysis page on the ImageJ wiki is another good read.

As stated on the Coloc 2 page (READ THIS WHOLE PAGE - for sure !!) also on the wiki … “Deconvolution is a great way to restore images to give a better estimate of the real fluorophore spatial distribution by suppressing noise and removing the offset or background…” Manders/Pearsons are really sensitive to non-zero background - so applying a mask/ROI is also helpful.

Would you be able to post some images for us to test as well - along with any ROIs/masks you applied? That was we can have a look at your images (upload your original datasets) and test ourselves.

Hope this helps to at least get you started… for sure others here on the Forum may provide other/better insight (@chalkie666??).

eta :slight_smile:

Hello, Thank you very much for your feedback. I am trying to become more familiar with the analysis but, unfortunately, not progressing as much as I want to.

First, I split the channels and only worked with the green and red channels, which correspond to my proteins of interest. I selected 3 ROIs in the image attached.

Unfortunately, I can not attach the image that I am working (ROIs) due to some error:“Maybe your image is corrupted.” So, I screenshot it for your consideration.

I have been following the Coloc2 page and also the “Deconvolution is a great way to restore images to give a better estimate of the real fluorophore spatial distribution by suppressing noise and removing the offset or background…” section.

I tried to use the mask tool as well and select the correspondent description to further process the analysis, but all the times I receive the same warning message. Please, see attached:


I believe the issue must be on my end to be very unfamiliar with this type of analysis, but I am very thankful for your suggestions to overcome this difficulty.

@GScola you need to take care of the non zero background offset in your images. Areas that should have zero or very close to zero have a larger value. You can measure it by measuring the average pixel value of a known empty area away from the cells. Then subtract that value from all.licrls in the image using subtract command.
Thus will get rid of the warning.
Persons values will be unaffected.
Manders will be affected.
Autothresholding will be strongly affected and more likely to do the right thing.

If this is from.a confocal you can avoid this extra step by setting the pmt offset voltage so that background areas have average pixel value between zero and 1.

If this is from a confocal watch out for noisy image data. You might want to deconvole (contrast restore) the images to fix systematic error of blur and remove some photon randomness noise at the same time. At the very least you might want to do a gentle gaussian smooth filter to suppress noise.

Noise is killer for the correlation methods used here.

Depending on the biological question, you might also want to exclude the areas covered by the cell nuclei. Only measure in places where the biology of interest is happening or the result is thrown off by irrelevant data.

Your images are mostly black background. This is a waste I’d disk space and scanning time.? especially when using a confocal. Set the scan area to cover one patch of cell at a time. Get lots and lots of images. You need to average over ma y many cells to get a good measurement.

Do you have a negative control? A positive control?
Without those anything measured might be nothing to do with the biology you are trying to measure.


Hello @chalkie666,

Thank you very much for all the explanation and help with this matter. I understand that we need to be very careful about the settings to acquire the biological information expected.

I used confocal microscopy to assess the colocalization. I have many other biochemical measurements that appoint the interaction between these 2 proteins; this is why I was expecting to acquire some colocalization to validate the biochemical data.

I do have negative and positive controls, and the majority of my images are from isolated cells. I am using that image as a training example. Perhaps I should select another field with only one cell.

I will follow your suggestions and provide you a feedback. Many thanks for your help and time on this analysis :slight_smile:


Sounds good!
Doesn’t matter how many cells there are in the field if view of the image, so long as you use a region of interest that excludes all parts of the image where the interesting biology is not happening. Eg. background and maybe also the nuclear area.

But more data in each image is a good thing.
Or simply lots of independent cell images.

From a stats perspective you might want to measure each cell as a individual data point then accumulate lots of those to get population statistics. Doesn’t mean only 1 cell per raw image. It means one cell per individual analysis. Just a thought.

Oh perfect! I am very new to this type of analysis. Very grateful for your guidance :slight_smile: