The control: a vital part of any scientific experiment. The only way to establish causality between the dependent and independent variables in an experiment is by comparison with a control. Controls are also important for establishing detection background: what does “nothing” look like on your sensor? Bioassays are no different: including control wells in your assay design is often a good idea. Not only do they provide the benefits one might expect from an experimental control, but they can also be exploited for accuracy and precision benefits in your results.
What is a control well?
In an experiment, a control is where we observe the effect of doing nothing. We replicate our set-up as closely as possible, with the only difference that we do not introduce the element we predict will cause an effect. A classic example in the biosciences is a placebo-controlled trial. We give one group of participants the treatment under investigation, and another – a placebo – which we expect to have no active pharmaceutical effect on their outcomes. By comparing the outcomes of those participants in the treatment group to those in the control group, we can deduce the effectiveness of the treatment.
In bioassays, we can introduce control wells into our plate layout. These can take a few different forms, which we will investigate in a moment. What they all have in common, however, is that they provide a baseline response in the assay. This is important for determining the amount of background noise present in the response data, as well as checking that the assay system is behaving as expected.
Types of control in bioassay
Broadly speaking, there are two main types of control which are typically used in bioassays: positive and negative.
Negative control wells are the closest to what we would consider a control in other contexts. Their purpose is to measure what is detected in the assay when the expected response is zero.
The simplest forms of a negative control are known as blank controls. These could simply be a completely empty well, or a well containing the assay buffer solution alone with no reagents or cells. These are intended to measure the background response due to the plastic of the cell and the buffer solution, and are expected to give a small response. If the observed response when the assay is unexpected, this could be an indication of a problem in the assay system.
Other forms of negative control contain more of the necessary reagents while still excluding any cells or analytes. Among the most extreme version of this type of negative control is the negative matrix control, which contains everything which would be included in a sample except for the components we would expect to elicit a response. This is intended to measure the so-called matrix effect: the background response due to the material in which a sample is contained. For example, a blood sample used in an ELISA is likely to contain components which scatter light used in reading the assay or otherwise affect the detected signal. These types of negative control give a more comprehensive understanding of the background response present in the assay, meaning it can be properly accounted for.
Sometimes, a positive control is also used. In development, the expected response generated from these positive controls can be determined, and then when the assay is in use, the positive control response seen can be compared with the expected response to determine if the assay is performing as intended.
An example might be a spiked matrix sample, where a known concentration of a sample of interest is added to the matrix which otherwise contains a very low concentration of the sample. This allows correlation of the expected response with the response actually observed in the assay. Other types of positive controls can be used to determine the maximal background response which could be observed, or even determine whether a particular matrix or buffer solution is suitable to be used in the assay at all.
Including control wells in an assay layout
As with any decision surrounding the design of a bioassay, the inclusion and location of control wells should be made strategically. Plate real estate is almost always limited, and the opportunity cost of including control wells over, say, another replicate can be high.
One solution to this is including control wells where one would be unlikely to locate a sample in the assay layout. For example, the edges of an assay plate are often avoided in plate layouts due to edge effects. Cells in wells at the outside edges of a bioassay plate can often grow differently to those in interior wells, leading to higher variability in results for the entire assay. There are several reasons why this might occur, but a solution is often to simply avoid plating samples using those wells.
Instead of simply wasting that plate real estate, however, it can often be prudent to use the edges as a location for control wells. As well as the efficiency benefits of making use of otherwise-unoccupied plate space, using carefully chosen control wells on the edges can give an indication of the magnitude of edge effects for a particular assay. This can inform decisions made about assay optimisation during development.
Using control well
As we’ve discussed, the main purpose of any control well is as a check that the assay system is performing as expected. If a response notably different from that expected is observed, then this could be a sign that there has been a problem in the assay which requires investigation.
In some circumstances, however, control wells can be used as an enhancement for your analysis itself. One example of this in particular uses a zero dose control. These are negative control wells which contain none of the sample of interest, and which, therefore, we would expect to give just the background response
To fit a statistical model to a dataset, we typically require at least one more response than there are free parameters in the model. So, to fit a 4PL, we require at least 5 data points, for example. It is not difficult to envision a scenario where a 4PL model would likely be the best fit for a data set, but an additional data point is required for the model to be fitted.
A solution could be to use the response from the control well as the extra data point! This would allow a 4PL to be fit instead of a linear model, providing an improved model fit for the data. This would likely reduce the likelihood of suitability failures on parallelism or goodness-of-fit.
This is not the same as using a 3PL model, however. In that model, one of the two asymptotes takes on an arbitrary fixed value which does not take into account assay-wide shifts in response. Since the control well is plated along with the samples in the assay, we would expect that any shifts would be reflected in the response seen from the control.
It should here be noted that not all statistical software packages can cope with including a zero-dose point. This is because the dose scale is typically represented on the log scale, and the logarithm of zero tends to negative infinity. The good news is that QuBAS, the bioassay statistical analysis package developed by Quantics, is built with the option to include a zero-dose control point with the push of a button. Even if you have enough data points to fit a model, it can be a good idea to include a zero-dose control point to give a better estimate of the zero-dose asymptote and better precision for reportable results.
Comments are closed.