2
$\begingroup$

I am using ImageJ to analyze Western Blots. I have scanned films in as grayscale images because this is how we did it in my old lab. People in my current lab are not satisfied with that explanation and think I should consider using color images. I've been searching for protocols but they all seem to address how to use the program once you have some films scanned.

Can anyone help me figure out what considerations go into choosing how I scan my films?

$\endgroup$
2
  • 2
    $\begingroup$ You are using film, so the resulting image is black-white anyway, right? ImageJ can of course also deal with grayscale images. $\endgroup$ – Chris Oct 1 '14 at 22:02
  • 1
    $\begingroup$ I've never used films in imageJ, but I have done fluorescent microscopy images, which come as black and white images, 1 image per channel. Then I have to go and set each image to a particular color channel, red, green, blue, etc, to create the final colored image. I don't know how to get colored images of western blots since most gel imaging cameras I've used have a black and white camera with multiple filters, creating a situation similar to the fluorescent microscopy images I've worked with. Of course I have no clue how film and scanners would work with this. $\endgroup$ – user137 Oct 1 '14 at 22:07
3
$\begingroup$

There is no evidence that one is better than the other, most likely because it differs from case to case. Neither you, nor your critics, are right. There is a tiny bit of science in a paper on digitizing blots, generalizing from blots of a specific protein (PMID: 19517440), and they use grayscale for no given reason. Come to think, that is the best paper in the field of immunoblot quantification., and it still lacks evidence.

Immunoblot is semiquantitative, meaning it may sort and order bands (and protein amounts), but it fails often at estimating differences and ratios, and may even fail to see differences. Just try your best to find a difference, in the knowledge that if there is no difference in protein amounts, any fiddling with contrast won't create a difference in bands quantitation.

There is almost no difference between color and black-and-white digitized images of an immunoblot. The parts of the blot that are black will still appear in the computer as White = 0, even though a tri-color image file will detail it to Red = 0, Green = 0, Blue = 0. The parts that are nearly black will still be very close to zero.

The only thing that changes a lot between RGB and color is the background. The background may change, for example, from White = 200, to Red = 190, Green = 190, Blue = 220. There is one consequence of such switching between values for background. (These changes may be induced by switching between grayscale and RGB, but also by altering exposure time etc.) When you use a method where the background gets too close to the blots, the "amplitude" of the blots, the height of the peaks in ImageJ, will be reduced. This loss of contrast should wash out some of the differences between blot bands and lessen support for differences between bands (i.e., for the alternative hypothesis), making you miss a genuine protein difference.

Again, if your blot bands are different enough so that your percent increase / decrease is huge, and your p value is small, you are all set, regardless of the method you chose. Ask your critics to explain your observed differences other than by differences in protein. But if your quantification fails the p test, I think it will be a honest effort, to change the color from grayscale to RGB, or the other way around, just in case your first choice had been the one that dampens contrast.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.