So, is Design of Experiments (DoE), just a passing trend? DoE might appear to be a relatively new buzz word or concept, but rest assured it’s worth has been proven long ago within other industries. Why therefore hasn’t the approach been readily adopted within the Life Sciences?
Focussing specifically on its use within assay development, there have been a few barriers to adoption. Firstly, integral to the concept of DoE is randomization. Whilst randomizing wells is entirely possible using the tool of choice for many assay development scientists, the hand-pipette, the more complex the experimental set-up, the easier it is for error to creep in!
The obvious option here is to seek the assistance of automation. Automation also opens the possibility of developing assays faster, in higher density plates that minimize reagent usage.
Transferring assay development into higher density plates using the basic liquid handlers routinely found in assay dev labs, however, represents another bottleneck. These liquid handlers tend to be fixed head robots not able to dispense into one well at a time and as such, are not set up to meet the demands of DoE. More sophisticated automated liquid handling solutions offer a feasible solution however, programming nightmares often ensue when working with different reagent classes in order to ensure accurate and precise dispensing. Unfortunately, fully understanding the intricacies of robot programming isn’t a skill set commonly found within assay development laboratories and complex liquid handling solutions can be left to gather dust.
If you have made your way through all the above, you might find yourself bottlenecked when it comes to making sense of the data mountain! As scientists we all have a basic understanding of statistics but when it comes to the statistical complexities of DoE, this is enough to strike the fear of a higher being into most of us.
Those pioneering the use of DoE for assay development purposes use generic statistic packages utilized by other industries to design experiments and analyze results, but these don’t integrate seamlessly with Life Science liquid handlers. The work around here has been to use manual spread sheets to map the experimental design out to the liquid handler then map data back for analysis, however this approach is cumbersome and poses a data integrity concern.
So, it’s fairly easy to understand why DoE hasn’t been readily adopted, however, the benefits of this method are undeniably advantageous over the traditional One Factor At a Time (OFAT) approach as discussed in our previous blog. However, what if there was a solution to eliminate these bottlenecks. Would you be willing to adopt?