Fig. 5
From: RefCell: multi-dimensional analysis of image-based high-throughput screens based on ‘typical cells’
The shift of mean cell properties by siRNA perturbations for classification boundaries computed from (a) typical cells and (b) randomly selected cells. Each green or red point represents the mean of all cells in one GFP-progerin repressed (healthy-like) or GFP-progerin expressing (progeria-like) control sample, respectively. There are 12 samples for each control type. Each blue point represents the mean of all cells for one siRNA perturbation. The classification boundary is shown as a vertical dotted black line. Four siRNA samples that deviate significantly from both controls in each of the four channels are labeled (siPHF13 for progerin; siNEDD4 for lamin B1; siTRIML1 for DAPI (nuclear shape), and siRNF8 for γH2AX). Note that the range of the x-axis is the same as the range of the y-axis in all panels. a Most points are preferentially shifted perpendicular to the classification boundary. Variation parallel to the classification boundary is small compared to the variation perpendicular to it. b siRNA perturbations are shifted both parallel and perpendicular to the classification boundary when the classification boundary is computed from randomly selected cells