hi there. I am in a major hurdle block and I hope to get advise on this. My experiment involves putting mechanical stress on cells at different durations and looking at whether a particular pathway is activated. So I look at phosphor levels of certain proteins. using western blot, I probe for the phospho protein first and expose them, before I stripped the antibody and use an antibody to detect the total protein. I then find the ratio of phosphor to total proteins, before normalizing this ratio against the control ratio.
I just do not know which values to input in and what statistical analysis to use. an example of the data is below attached, there are 2 other replicates.
can anyone advise on what statistical analysis to use and which values to input? I want to compare from the 3 runs whether each duration is significantly different than control
As you have 5 runs first I would put all the 5 values (time by time) and compare with the 5 values of the control. It doe snot matter whether you use "ratio" or normalized.
But there may be a "big problem" that should be solved before doing any statistics, which is how the p38 may be higher than the total p38. Either I miss something or this is impossible (ratios cannot be higher than 1).
Agree with Rafael on the total p38 vs. pp38. I suspect the pp38 antibody is more robust in your assays. You should normalize both signals to a pair of standards first before running any statistical analysis. This will also reduce experiment to experiment variations.
A key problem you're going to run into here is this: Western blots are not quantitative. They simply are not. They are semi-quantitative at best, and this makes statistical analysis of their results... hazardous. Similarly, you get different answers based upon how long you expose your film (or keep your camera open, or whatever technique you use to record data). A 5 second exposure will give a different ratio of spot densities than a 10 second exposure. Similarly, a 5 second exposure on one membrane will give different results than a 5 second exposure on a second membrane - it's very nearly impossible to get identical antibody staining twice in a row, let alone 5 times. And stripping your membrane is not an ideal way to go for total protein - you remove some protein during stripping (that's the whole point, after all, is to get rid of protein), and it tends to be removed more from the lanes with more protein (ie: the procedure changes the relative amounts of protein from one lane to another). But if you run two individual gels, you run the risk of slightly different loading from one gel to the next.
You'd be better off using a more quantitative technique if you want to determine statistical significance. But if you're really determined to base it on Western blots: get a third blot, for some control band. Actin or vimentin or vinculin or some other thing you don't expect to change. Normalize your p-p38 to that: divide each band's density by the density of the related actin band. Then scale the bands to the control band: Control becomes 1, everything else is some fraction of that. Now do the same for p38. This eliminates the problem of one antibody or exposure being darker than the next, and gives you numbers of the same scale. NOW determine your ratios, using just the scaled numbers. To determine p-values, it depends upon what you want to show. Do you want to show the amount of p-p38 is different? Base it on p-p38 readings. Do you want to show the ratio is different? Base it on the calculated ratio.
- Badges
- Science topic