Are you using the Analyze>Gels tool (http://imagej.nih.gov/ij/docs/guide/146-30.html#sec:Analyze-Menu section 30.11)? If so you should select one rectangle per lane and not all the lanes with the same rectangle as it looks by the images. This makes peaks smother and also separated (specially if the WB is clean and there is only one peak per lane). Doing this tou´ll end up with peaks with a gaussian-like distribution and kwoning this makes a lot easier to decide how to mark individual peaks.

Here you can find an excelent video tutorial:

http://imagejdocu.tudor.lu/doku.php?id=video:analysis:gel_quantification_analysis

However, to substract the background instead of the Substract Background Tool used in the video I prefer to select an empty lane which gives me the basal gray level in the blot.

Finally, even more important than this, it is how you digitalize the WBs. In the link to the Analyze>Gels tool you´ll find how to do it right (i not saying that you are doing it wrong now but just in case check it out).

Cheers.

I agree with Victor. The best way is to select an empty lane where no protein was loaded and take it as background. This number should be subtracted from each lane signal.

We always measure the area of the spot (each active (target) spot). Sometimes background is uneven, than you need measure it as area close to your each active spot.... and substracting. If the background is good, a few measure is enough on different small parts on the digi pic and that case, it is similar usually (the average value of pixel density). This is the Measure-Area tool i think. The line analysis should not give this type of signal... if we speak about western. Did you measure it cross-line, mighbe?

Thank you you All

Dear Victor, Yes, I used Analyze>Gels tool in ImagJ.

Dear Attila, Cross-line? If you meant I select all lanes in one rectangle. then yes.  May I ask where is this Measure-Area tool that you mentioned in Imagej?

A few question I'd like to add:

1.  Is there a way to move select next lane horizontally? When the band is wider twice than height, the tool won't allow me to move horizontally, only vertically is allowed.

2. You guys mentioned to select empty lane as background value, and subtract this number from the band. How do you this? Do you mean measuring the area and then subtract it from the band value in excel? If so, I am little confused.

2.1) How do you select area of the plot lane? It comes up like many small moutains/peaks.

2.2)  in the video guide, they draw a line under the peak to set the background. This mean the background has been subtracted from the band. Why do you need to measure empty lane for background?

3. Can I just adjust the image's brightness until background totally white. This will make the peak starting and ending points drop at x axis =0

Thank you

Dear Asa,

1) In a blot or any other gel, the lane height it is always longer than wider. And you need to draw the selecction that way. No matter if there is only one band o many bands per lane.

2) Did you read the ImageJ guide? You just need to select the area under the peak with the magic band tool after closing the peaks with a line.

2.2 and 3) Yes in the video background has been substracted, but if the background it is uniform I try to avoid this tool. Selecting an empty lane give you baseline that you can use as reference to draw the lane under the other peaks. Also by selecting the whole lane and not only a region around the band you are including a background region around the peak that should gave more or less the same height as the empty lane (if there is not empty lane you can use it as baseline to close the peaks). You do not need (and should not) modifiy brigtness in a image that you want to quantify. It is better to use the Raw file to quantitate.

Hm, i see. we measure spots individually. Select one with rectangle or free-hand tool, than use Analyze ->Measure command. U can set some parameters in the Analyze-> Set Measure... part. Ctr + M is the hot key for measuring. You will get the result a small table can be copied to other programs.

Thank you Victor. Let me summarize what I understand from what you wrttien if it is correct or not:

To select empty lane is just for giving a guide/idea to how to draw a baseline to closure other band peak. No need to measure empty lane area.

*At first I thought to use wand tool to measure empty lane area too, then use this value to minus value of the band in excel. I tried this. The problem is I am not sure how can I closure area for the empty lane , the plot comes up as many peak. (attached pic)

Tha´s correct. I´ll do it more or less like this.

And remember that this method it is only an fast&cheap alternative to directly quantify the blot using a Chemidoc or similar and it is not perfect.

I also agree with Victor. The best way could be to select an empty lane where no protein was loaded and utilize it as background. This value should be subtracted from each calculated value.

I do not agree, using ECL lead to not homogeneous background which has multiple explanations (diffusion, transfer, not homogeneous incubation/shaking).

Best case, method in fig2: remove to the high signal lot of background, in low signal less. THIS IS VERY CONSERVATIVE as you are decreasing differences between groups.

Worst case, method in the fig1: considering homogeneous background is definitely wrong. Cutting signal in some peak is destructive and will result to high variation between experiments.

Concerning rolling ball: it looks simpler and safer but it is not reliable (http://www.ncbi.nlm.nih.gov/pubmed/19517440).

Therefore, I just remove a bit of background by removing the intensity value of the lowest pixel to all pixel, that increase a bit my dynamic range and does not change my raw data. That said, I do not consider a blot with a large variation of the background (need long exposure to evaluate it).

I know that more advanced methods are used by many well developed quantification software, but I rather redo a blot that rely on these algorithms:

1/ one band: background in the complete lane (except the band of interest) is measured and made relative per pixel.

2/ multiple bands: too complex to measure the lane background. Focus on the signal in between the lane of interest and the left-right lane following.

I completly agree with everything that Nicolas says but I do not think he is right because the question is not about the best way of quantitate a WB, but the best way of doing it using ImageJ, which is a very nice freeware for image analysis but not well suited for gel analysis. However, for everyday use it´s covenient and people uses it a lot for this. If it´s better to quantitate a scanned (and ,as I mentioned above, this is by no means trivial) image of a film which may be saturated or to directly quantitate photons from the ECL with an appropiate system it is another question.

But if Asa is analyzing WB with ImageJ he must select lane individualy and not use the same ROI for the same band in different lanes. As I show in the attached images if you open the Sample Gel in imageJ (Flie>Open Samples> Gel) when you select all the band of the same height in a single region the peaks are to close and it is imposible to separate then. Even worse there is not peak in the first lane. But if you follow the ImageJ manual and use different regions for different lanes you can usually separate the peaks well enough. I only show the first two lanes but you can see (in green) two peaks corresponding to the top band of the two first lanes, and the more intense band has clearly more area which at the end it is what ImageJ uses with the tool of gel analysis. With this tool it is not possible to select regions between lane unless you left lanes unloades in the gel. That´s why I like to select an empty lane to give me an idea of the background (of course if the background is not even then you must repeat the blot). It will also work a zone of background of the lane selected but if there are many bands (as you can see in the image) the peaks overlap (tha´t why I use a region quite far away).

With everythinh else I agree...

I have to agree with Nicolas - the fundamental issue here is that Western blotting is at best a semi-quantitative method. The different background subtraction methods will never add to the accuracy of the results, they will only limit the effect of different approximations.

On a clean blot, the three different methods will give you very similar results (certainly within experimental error), but on a dirty blot the question is, where is the background coming from? If you don’t know that, it is difficult to predict which method will introduce the fewest errors.

Victor pointed out that the distribution of each band should be considered Gaussian, but ImageJ doesn’t have a tool to model the background as the sum of multiple Gaussians, which is the inherent problem of quantifying each lane. Doing the horizontal method, as Asa did, reduces but does not elimiinate this problem.

In short, wotk on optimizing your blot. Then you won’t have to worry about your background method.

Thank you Victor for moderating the topic.

I do not want to discredit imageJ and western blot, just want to warn on other issues which might be tentative to ignore and which will result to problem in mid term.

Bu you are absolutely right, picking as you describe the ROI will do the job relatively well. I like a lot the example you are using to illustrate your concept.

Thank you Nicolas

May I ask you guys  what software do you usually use for western blot data analysis?

Ok, then. It looks we all agree. There a lot of issues to be concerned when you try to quantitate a blot. But my point was that it is crucial to use correctly the software if you want to get a chance to get it right.

Asa,  I recommend users at my facility ImajeJ/Fiji a lot, but I also advise them about how to scan films correctly. We also have an Odyssey Imaging System at my institution to directly quantitate the membrane. I believe that this (or a similar system from other brand) it is more accurate because it has a good dinamyc range and it allows you to work in a linear range where the protein concentration in the gel is directly proportional to the signal. With film, this is not the case if the bands are too dark.

good luck with your research Asa.

Software to use: Image lab and ChemiDoc from BioRad. For analysis with Chemi Doc I scan X-ray pictures with common scanner (hp scanjet 5590). Currently I moved to ChemiDoc(no X-Ray film needed) and I am making all analysis with this application - to obtain the raw data. But the density values (volumes) I calculate basically in Excel after importing data from ChemiDoc. Here you have full controll of what you are doing. Important thing is to have pictures on which film is not saturated. In above example bands are (looks like) oversaturated and for that reason are not very suitable to be used for quantification. And background (reference) for given band I select for each lane separately.

There is a nice article looking at densitometry, using different quantification softwares and techniques, and then comparing the results with values obtained from radioimmunoassays. Gassmann et al. "Quantifying Western blots: Pitfalls of densitometry."  Electrophoresis 2009.

If you use "subtract background" in ImageJ, all your peaks should have the same baseline.