How to use it
- Measure the target band and the loading control band in ImageJ. Subtract the background there, before you come here.
- Paste one line per lane: the lane name, the target density and the loading control density.
- To fold replicates together, add a group name as a fourth column.
- Pick the control group. Every fold change is measured with that group's mean set to 1.
- Read the step 1 and step 2 columns together. The column beside them is what you would have got by skipping step 1.
The formulas, and what to watch for
The arithmetic is two divisions. The hard part is not the arithmetic, it is the order.
Step 1 — divide each lane by its own loading control
Lanes never carry the same amount of protein, so the target density is divided by the loading control density of that same lane. The result is what the target would have been if the loading had been equal.
- Normalised density
- target ÷ loading control
Step 2 — divide by the control group's mean
The step 1 value is divided by the mean of the control lanes' step 1 values. The control becomes 1 and everything else is a fold change against it.
- Fold change
- normalised ÷ mean normalised of control
What to watch for
- The source writes step 1 as target × (control lane's loading ÷ this lane's loading). Here it is written as target ÷ this lane's loading. The two differ by one constant — the control lane's loading control — which step 2 divides out of every lane alike, so the fold changes are identical.
- Densities must be background-subtracted, and measured from an unsaturated image. A saturated band shows a smaller difference than the real one.
- The spread of the loading controls is reported as a CV, with no pass or fail line drawn on it. No primary document stating such a threshold was found, and this site does not put an unsourced number where a criterion goes.
- Housekeeping proteins have typically saturated the membrane above roughly 4 μg per lane (Taylor & Posch, 2014). A saturated loading control looks the same however much was loaded, so normalising against it erases the very difference you are measuring.
- The unnormalised column is for comparison and is not a number to report. It is there to show how far off step 1 would have left you.
Frequently asked questions
Yes. Dividing by the control first and by the loading control afterwards divides each lane by a different constant, and the result moves. The unnormalised column shows what skipping step 1 entirely would have given, so you can see the size of it on your own data.
Yes, and the arithmetic is unchanged — put the total protein signal in the third cell. Total protein is reported to be less prone to the saturation problem that housekeeping proteins run into.
Their step 1 values are averaged and that average is the reference. Individual control lanes therefore come out near 1 rather than exactly 1, and that scatter is the reproducibility of the experiment.
A background left in inflates the weaker bands proportionally more. This tool cannot tell whether it was subtracted, so it cannot warn you — do it in ImageJ first.