A Design of Experiment Approach to Optimize an Image Analysis Protocol for Drug Screening

The Design of Experiment, DoE, was applied to support the development of an innovative optical platform for ion channel drug screening. In this work, DoE was exploited to investigate a set of software parameters instead of process variables, an approach that has been only rarely explored. In particular, it was used to define a standard analytical configuration for a MatLab-based image analysis software that has been developed in the laboratory to extract information from images acquired under the drug screening conditions. Since the choice of the type of analysis and filtering, as well as their interactions, was known to affect the final result, the aim was to identify a robust set of conditions in order to obtain reliable concentration-response (sigmoidal) curves in an automated way. We considered five parameters as factors (all qualitative) and two characteristics of the sigmoidal curve as reference outputs. A first DoE screening was performed to reduce the number of needed levels for one factor (an unconventional approach) and a second optimization study to define the best configuration setting. Image stacks from three different experimentation days were used for the analysis and modelled by blocks to investigate inter-day variations. The optimized set of parameters identified in this way was successfully validated on different cell lines exposed to their references drugs. Thanks to this study, we were able to: find the optimized configuration for the analysis, with a reduced number of trials compared to the classical “One Variable at A Time” approach; acquire information about the interactions of different analytical conditions as well as the inter-day influence; and, finally, obtain statistical evidence to make results more robust.