小中大Data analysis
1. Once you have measured the values for all of your bands, enter your numbers in a spreadsheet. Make a list of your samples on the blot, and then two adjacent columns of the Mean and Pixel values for each sample's band. (Note, if you used Image J's Gel Analysis routine, simply paste in the percent values for each sample)
2. Multiply the Mean value by the Pixel value for each band. This gives us an integrated measure of the intensity and size of the band. I'll refer to this as the absolute intensity. (Note: if you used ImageJ's Gel Analysis routine, this step does not apply)
3. Next we'll calculate a Relative Intensity, using our standard as the common point of comparison. Divide the absolute intensity of each sample band by the absolute intensity of your standard to come up with a Relative Intensity for each sample band. Some bands will have a Relative Intensity lower than 1 (they have less protein than the standard), and some bands might have a Relative Intensity larger than 1 (they have more protein than the standard). The Relative Intensity is a unitless value. (Note: if you used ImageJ's Gel Analysis routine, divide the percent value for each sample by the percent value for the standard from that membrane to get a value equivalent to Relative Intensity)
4. If you have the same sample standard on multiple membranes, you can compare intensity values across multiple membranes, even if you had to expose them for different times. By calculating a relative intensity that is tied to the same sample standard on every membrane (10ng of Human Hsp70 for example), we can make up for variations in the length of film exposure or variations in the efficiency of the antibodies or other reagents.
5. In order to test for significant differences between treatments, all of your membranes will need to be scanned and quantified, then expressed in terms of Relative Intensity. If you are going to test for treatment effects using a standard analysis of variance, you will need to ensure that your Relative Intensity values are normally distributed and that there is homogeneity of variances within each treatment. A log transformation is often needed to make Relative Intensity values approximately normally distributed, but this may vary depending on your data. The complete statistical analysis of the data is outside the scope of this article, please consult a statistics textbook for more information.
6. For making figures, your data can be plotted as Relative Intensity versus the treatments, and most papers typically use the standard error of the mean for the error bars. It should be noted here that some researchers make the extra effort to include a set of serial dilutions of a known standard on each Western Blot. Using the serial dilution curve and the quantification techniques outlined above, it should be possible to express your sample bands in terms of picograms or nanograms of protein.
An example figure showing increased expression of a protein at high temperatures
Addendum
With regard the the Image-J gel analysis routine (method 2 above), there has been some question of what the values reported by Image-J correspond to. The images below may help illustrate what Image-J is measuring.
In the image above, I have drawn out a set of fake "bands" in Adobe Illustrator. The gray value and area of each band are listed above the band. Additionally, I have included the "area" value returned by Image-J after plotting the bands and clicking in each peak with the Wand tool. Note that these "area" values are a RELATIVE measure of the size and density of each peak you clicked with the wand tool. When you halve the area of a band, but maintain the same gray value (compare lanes 1+2), the value reported by Image-J is half as large. By the same token, if you halve the gray value but maintain the same area (compare lanes 1+5), the value reported by Image-J is halved.
The same holds true for bands of different shapes. In the image above, altering the shape of the band, but maintaining the same gray value and area (compare lanes 1+3) yields an equivalent value from Image-J.
Image-J also accounts for gaps in a band, as shown above. Compare lanes 1+3, which both have an equal number of gray pixels and equal gray values (i.e. equal amounts of protein on the gel). Image-J reports the same "area" value for both of these lanes.
It is worth reiterating that the "area" values and percentages reported by Image-J are always relative to the total size and density of bands that you have selected in a particular image. In the image immediately above, the band in column 1 returns an "area" value of 4000, while in the previous two images column 1 had the same size band, but with twice the gray value, which in both cases also returned a value of 4000. The raw values returned by Image-J are meaningless for comparing across different gels, since they are only a relative measure of the bands you've highlighted on a particular gel image. This is why we need to standardize to some common standard loaded onto all of the gels.
(Edit July 2009) This may be a bad sign for my career, but it seems that this set of instructions is quickly on its way to becoming my most heavily cited publication in the scientific literature (see my "real" publications here (pdf) ).
For instance:
Corpas F. J., et al. 2008. Metabolism of Reactive Nitrogen Species in Pea Plants Under Abiotic Stress Conditions. Plant Cell Physiology, 49(11): 1711-1722
Luhtala N. & Parker, R, 2009. LSM1 over-expression in Saccharomyces cerevisiae depletes U6 snRNA levels. Nucleic Acids Research, doi: 10.1093/nar/gkp572
Miller, R. K. et al. 2009. CSN-5, a Component of the COP9 Signalosome Complex, Regulates the Levels of UNC-96 and UNC-98, two Components of M-lines in C. elegans Muscle. Molecular Biology of the Cell
Chiang, E. T. et al, 2009. Protective effects of high-molecular weight Polyethylene Glycol (PEG) in human lung endothelial cell barrier regulation: Role of actin cytoskeletal rearrangement. Microvascular Research 77(2): 174-186