Hi, I have been working on analysing a series of drug screens in 384-well plates. The screen is split over 4x384 well plates of ~850 drugs all assayed at 1uM concentration. There are multiple biological replicates but not technical replicates.
An example design of the plate is shown below:
For analysis, I have currently been using two measures:
Percentage of control (POC)
Robust z-score calculated for each plate separately (only using drug wells, excluding controls)
The issue that I have when comparing between plates and also between replicates is that using the robust Z-score gives me plate specific differences:
Different plates are shown here in different colours
I firstly wanted to double check that calculating the robust Z-score per plate and makes sense and whether I had to do anything else to 'normalise' out the plate-specific effect. I had assumed the point of doing the Z-score was to remove the plate specific effects.
Secondly, I wanted to ask for advice on any other better approach to analysing this kind of data. Being a single concentration and the plate design being as it is, this kind of data isn't compatible with pipelines like cellHTS2.
Any input is appreciated.