Data Analysis Pipeline for High Content Screening in Drug Discovery - High-Throughput Image Reconstruction and Analysis - page 222

Figure 9.7 Support vector machine mitotic subphase classification. (a) A supervised support vector
machine (SVM) model was implemented to generate a multidimensional classification filter based
on expert annotated images of cells in mitotic subphases. (b) Overall sensitivity and specificity of
the method were above 90%.

The percentage of cells in each phase was then calculated for each well. Typ-
ically, most cells are in interphase (G1, S, or G2), with only a small number of
cells in M-phase. Therefore, the percentages of cells in each cell cycle phase vary
significantly among the phases and need to be standardized before the comparison
of cell cycle profiles of different treatments can be made.
We introduced the NZ score for this purpose. If the percentage for well i is
x i ,where i

= ...

N and N is the total number of wells in the reagent set, then the
NZ score was calculated as follows:

1

m − x i

NZ i =

· s

1.4826