# What is the best way from remove background signal from a band when doing Western blot image analysis?

I am doing image analysis of a Western Blot in Image J. I have calculated the total intensity of my protein bands of interest through outlining each band using the rectangle tool in ImageJ, and calculating the area and mean intensity. I have multiplied the area with the mean intensity to obtain the total intensity. However, I want to remove the background signal from each protein band of interest.

The background adjusted protein signal would be calculated as: (protein band mean intensity - background band mean intensity) * (area of protein band). On my blot, I have two isoforms of my protein of interest that were detected by the antibody, and I want to subtract the background from all bands for both proteins.

The way I am calculating the background, is for each lane, making a rectangular selection near the top of the lane where there is no labelling against any protein. The length of the rectangular section is equal to the length of the lane. However I am not sure how wide the rectangle for the background should be. From what I have seen online, there is no fixed method for calculating the background signal.

Are there any ways to determine how wide you want the band for your background to be? E.g. should the width of the band for the background be equal to the width of the protein band of interest, or can it be wider than that?

Additionally, how can you know if you have made the rectangular band for the background too wide?

Any insights are appreciated.

I haven't done immunoblots for years, and didn't do much quantitative analysis of them even then, so caveat emptor. However:

1) "Are there any ways to determine how wide you want the band for your background to be?"

I would say that the central limit theorem is a good place to start. Basically, as the size of your background increases, the precision of your estimate of mean background intensity will also increase (asymptotically). So you want your background sample to be as big as possible.

This does have some assumptions:

1. Little regional fluctuation of mean background signal

2. (related) Absence of gel artifacts

So, you want your background sample to be as big as it can be without including any artifacts and as long as it is representative of the blot in the general region of your band. In practice, I imagine that taking a selection the size of your protein band is probably fine in most cases, and probably easier to explain. You just have to pick a representative piece of background to measure.

2) how can you know if you have made the rectangular band for the background too wide?

Wider than the blot is too wide. On an infinitely large blot, I think you can't make it too wide. For practical purposes, you want it to be as wide as you can get it and not overlap obvious artifacts or the edge of the blot. But again, it's probably easier to explain just selecting a piece of background as big as the protein band that is roughly near the protein band, to account for local background variation.