I’ve been in science and thinking about science for long enough that there are a lot of concepts and terms that are second nature to me that are completely foreign to most of the public (a friend recently brought up the word “assay” as an example of this: I use it all the time and it has a specific meaning, but for the general public it’s a less common word that’s sort of just synonymous with “experiment” or “test”. Or it has a slightly different industry-specific meaning).
But today I want to talk about controls. This may be really basic, but I just want to get everyone on the same page quickly. If you change a variable in an experiment, it means nothing if you don’t have a baseline to compare to. Say we want to study the effect of drug A on patients. To figure out what effects drug A has, we have to have another group of patients who get a control, which will be our baseline with which we can compare drug A patients.
And again, this may be basic, but bear with me. No drug is a terrible control. That’s why we use placebos. To separate out the effects of the act of giving someone a pill, any pill, and the effect of drug A itself.
Controls are some of the most important parts of an experiment because they are what determine how your data can or should be interpreted. And many experiments, especially as they become more complicated, actually need multiple controls. You need to confirm not only whether or not your experimental condition had an effect, but also whether that effect caused by what you think it was, and whether your experimental condition was actually present in the first place.
That last part is pretty easy in the drug A example (you either gave patients that drug or not, though even then, there’s a question of whether patients absorbed the drug and whether it got to the tissues you’d expect it to). But oftentimes in biology experiments, you actually don’t know if you did what you thought you did because you can’t see it. So you have to do further experiments and controls to say, “Yes, the cells are definitely there, and we’re not just seeing an effect of no fluorescence because all the cells died.”
And we talk a lot about controls in grad school and classes and science in general (that’s often the quickest critique of published research is what controls they did and whether those were sufficient to reasonably believe the authors’ conclusions).
But there’s one aspect that I think doesn’t get talked about enough, which is verification of results via other methods. Whenever you have a question in science, there are typically several ways of approaching it to try to get an answer. And too often (not all the time, but too often), not enough approaches are taken. Unfortunately, repeating research isn’t rewarded in science (it’s difficult to publish and impossible to get a grant for, so there’s very little incentive to confirm previous findings).
So how does this apply to me? I’m currently working on one experiment, but it has three parts. I’ve mentioned before that I’m injecting frog embryos with a morpholino (which prevents production of a certain protein). So I end up with embryos that haven’t made enough of the protein during development. And of course I have a control which was injected with a morpholino that doesn’t have a target (so I know that any effect I see isn’t simply because of the physical injection itself).
But the other two parts are actually verifying that my morpholino did what it was supposed to in the first place. I mentioned immunocytochemistry in my post two weeks ago, and that’s one approach. If I successfully prevented protein production, I would expect to see less fluorescence in neurons from my target protein morpholino embryos than from the control morpholino embryos. And I’m also doing a technique called a Western Blot to look at the relative concentrations of my target protein in each group. If both approaches agree, then I can be quite confident that my morpholino really did work, which increases the chances that any difference I see between my experimental and control embryos are actually due to the lack of the target protein in the experimental embryos. But if they disagree, then I might have a problem (or I might not). But it means I need to go back and look at my experiment to figure out why these two different approaches are telling me different things.
Similarly, to be confident in our scientific conclusions, it’s also good to approach the same larger problem from a few directions. So once I am confident that my morpholino is having an effect, I’m going to try another method of reducing this protein. Instead of preventing the production of the protein, I’m going to use another technique that will break this protein after it’s made to prevent it from doing it’s job. If both approaches agree, I can be very confident that this protein is necessary for the neuron to do X (where X is the function that breaks or changes when I get rid of the protein).