How Design of Experiments (DOE) boosts productivity
Bioanalytical laboratories often need to quickly develop robust immunoassays to support studies with tight timelines. Immunoassay performance is governed by many factors, such as reagent concentrations and sample dilution, and the classic experimental approach to improving assay performance involves changing one factor at a time. This approach is time consuming, misses important interactions between factors, and seldom leads to an optimal assay. There is an alternative, very powerful approach that has been used in many industries and applications to optimize processes – Design of Experiments (DOE). DOE has been successfully applied to immunoassay development by a number of research groups, including those using Gyrolab technology.
Missing the optimum is a real danger
Let’s look at a common challenge in an immunoassay – determining the concentration of capture and detection reagents needed to maximize signal/background and minimize variability. In the One Factor at a Time (OFT) approach, the first step could involve keeping the detection reagent at a fixed level based on experience and increasing the capture reagent successively (see Figure 1). At one point, the error reaches a local minimum. We could fix the concentration of the capture reagent at this level and then vary the concentration of the detection reagent. Again, increasing the level of the detection reagent could decrease variability further, entering a seemingly stable plateau of ‘optimum’ low variability before performance drops off again. We could conclude that we reached an optimum at ‘little X’ (see Figure 1). But we have completely missed the true optimum (big ‘X’) since we did not examine the interaction of the two factors, as seen in the contour map (or ‘response surface’) shown in Figure 1.
Create a model to build on
DOE software uses the input from your knowledge of the immunoassay (relevant factors and concentration ranges) to design a series of experiments in which all factors are independent of each other despite being varied simultaneously. In the case of optimizing capture and detection reagents, upper and lower levels are set for each reagent and all combinations are tested, together with triplicates in the middle for statistical purposes, making seven experiments in total (see Figure 2). The result is a model that opens a window on the contour plot and indicates the way towards the true optimum, which can be reached with additional experiments suggested by the software. DOE software also enables you to minimize the number of experiments required to a practical level if many factors are involved.
Figure 1: Changing one factor at a time may easily miss the true optimum (big ‘X’)
Figure 2: Using DOE, only seven experiments are needed to open a window onto the contour map that shows the way to the true optimum (X).
A model with many uses
The model built by DOE can then be used for four main steps:
Screening – determine which factors influence the responses, and over which ranges.
Optimization – determine the levels for key factors to obtain optimal responses.
Robustness testing – determine if the optimal settings of the factors actually result in a robust assay, or if they need to be adjusted.
Troubleshooting – the model can help in determining the cause of problems in assay performance
Gyrolab system and DOE make a good fit
Gyrolab system has proven itself time and time again as a powerful tool for rapid development of robust immunoassays. Several researchers have raised this high productivity to the next level by applying DOE to the design of Gyrolab assays. For example, assay developers at Pfizer have applied DOE to streamline the development of Gyrolab assays to achieve high robustness, in-study predictability, and improve the reliability of results. You can find out more by downloading the case study.
Find out more