Can you believe, I learned something useful to my studies of polymerization from my social amoeba experiments in your lab? Alona Bozhchenko asked with her big warm smile, at lunch yesterday at Wiess College. I came up with the most amazing controls, things that surprised my intern partners, she said. No, of course they were not the same controls as her social amoeba experiments needed. But all the time we spent agonizing over the right controls taught Alona in a visceral way that controls were important, that an experiment without the right controls would need to be repeated, and that the right controls were not always obvious. Any time you remember to think hard about something before doing it, the outcome will improve.
Alona worked as a technician in our lab for the year she took off from Rice. Now she’s back full time, taking more than a full load of chemical engineering courses, but she still had time to join us for lunch. Her summer internship gave her the experiences in polymerization and I was delighted to hear about them. You see, we always say that research in one area can enrich one’s ability to do research in any area, but examples of exactly why this is true are always welcome.
What exactly did she learn about controls from the year of often frustrating experiments with Dictyostelium? Alona’s experiment compared two different kinds of Dictyostelium clones. We wanted to know if one class of clones would cheat the other class during social interactions. In this case cheating meant increasing the proportion of spores produced at the end over the proportion of cells at the beginning. To test this, Alona and our other technician, Silven Read, put down mixtures of cells from log growth from two clones, and then allowed them to go through the social stage. This involves the amoebae aggregating, forming a slug, then a fruiting body in which about 20% of the cells die to form a hardy stalk that lifts the rest of the cells up. Comparing two classes of clones is a common experiment in our lab. We have done it for healthy vs. unhealthy cells, for example, and were now comparing farmers and non-farmers.
Would one class of clones cheat the others? What controls would we need? We labeled the cells with Cell Tracker, so one kind of control was whether or not the label itself had an effect. After all, our question was not about the label, but about the cells we labeled. If we always labeled the farmers and left the non-farmers unlabeled, we would not be able to distinguish the effect of the label from the effect of farming status.
We needed to be sure both clones in a mix were healthy, so we needed to plate each out clonally and be sure they formed normal and equivalent fruiting bodies. This is another kind of control.
A replicate is also a kind of control, though there are several different kinds of replicates. In an experiment like this one, we can do the whole thing over with the exact same clones, just to be sure there wasn’t some bias on a given day. When we do this, we never get the identical results, because we cannot do things absolutely the same, but we should get the same patterns.
In field work and many other kinds of research, exact replicates are not possible. If I remove the queen from a wasp nest to see who takes over, I cannot go back in time and remove that queen again. If I chop the head off a mouse to measure something, I cannot go back again and chop the head off a second time. With our social amoebae, however, we can go back to the freezer, grow up the identical clone, then subject it to the identical experiment.
But we also want to be sure that the result is not unique to a specific pair of clones, so we do more than one pair of genetically different clones. You see, we don’t want to know an answer specific to one pair of clones. We want to know if farmer clones as a class cheat non-farmer clones. This is a kind of control the wasp and mouse people can also do. They can study multiple wasp nests, and cut the heads off as many mice as needed. These controls will not be identical, but should differ as a class only in the treatment variable.
There are other aspects of experimental design that we were careful to teach Alona and Silven. You have to do experimentals and controls on the same day, for there will be effects from day. You can’t put all your experimentals on the top shelf of the incubator and all your controls on the bottom shelf. You can’t have one person collect the first half of the data and another collect the second half. After all, they may do things slightly differently in ways that could bias results. You need to worry all the time about these things.
When you think about confounding factors, there are two kinds to worry about. The less serious kind increases the variability in your experiments. It makes it harder to get significant results. The scary kind of problem introduces a systematic bias, that makes it look like you have a pattern when you don’t. We spend a lot of time worrying about how to design experiments in ways that are unbiased.
Why do newcomers to scientific experimentation have such a hard time with controls? I think it is because they think we are better than we are. Some people imagine that if we do something twice we would get exactly the same result. We don’t. We operate in a sea of noise. Only with all the right controls do we have any hope of detecting a meaningful signal above the chaos.
So, did those farmers cheat the non-farmers? Stay tuned for the very exciting paper coming soon to a peer-reviewed journal near you! As for Alona, her friendly brilliance and great sense of organization should help her to a fulfilling career, whether it involves plastics or amoebae.
The life cycles of social amoebae. Alona and Silven were studying the social cycle, at the bottom left. The other cycles are interesting too. David Brown and I devised this figure, and it is freely usable, subject to a Creative Commons 3.0 copyright, and found on dictybase.