Fig. 5: Light bleedthrough corrupts estimates of correlation and variance in single-cell intensities.
a Schematic representation of a cell (green) some distance from the centre of the microcolony, with its neighbouring cells labelled (lighter green). The intensity of individual cells depends on their position within the microcolony, given by dc/D, where dc is the distance from the colony centre, and D is the colony diameter, and the number of direct neighbours (cellN). b Cells closer to the centre of a simulated colony appear brighter than cells at the periphery due to light bleedthrough (Error bars = 99% CI, data sample averaged for colonies of size 20–1000 cells). The position-dependent trend predicted from simulated microcolonies (green) is consistent with experimental results (orange). c Simulated cells with more neighbours appear brighter (Error bars = 99% CI, data sample averaged for colonies of size 20–250) consistent with experiments (orange). d Two simulated colonies with CVs of 0.01 and 0.22, respectively, along with their convolution with the iPSF, showing that their observed CVs are very similar despite markedly different underlying cell intensity distributions. e At low noise, true CV < 0.15, where the underlying cell intensity distribution is uniform, the PSF causes artificial position-dependent variation in cell intensities (increasing CV). Conversely, when the input CV is high (true CV > 0.15), the PSF acts as a blurring filter, lowering the variance in the population by allocating intensities from brighter cells to neighbouring dimmer cells (lowering CV).
