Sketch of the characterization of clusters of infected cells.

<p>(<b>A</b>) Example of measured data in a grid of cells with the HCV RNA content per cell (left) and sketch of the ring structure of the cluster (right) as it would be defined by the algorithm shown in (<b>B</b>). Darker shading of cells indicates a higher amount of HCV RNA. The fitting procedure of the Matérn cluster process to estimate the domain radius accounts for edge effects due to sampling, i.e., only parts of the cluster might be visible on the grid of liver tissue analyzed by scLCM. The sketch in (<b>A</b>) shows an example for a cluster that grew spherically. The algorithm also allows for cluster growth that is skewed in one direction. (<b>B</b>) Example of the algorithm to determine the “ring structure” of a cluster of infected cells for a grid of cells. The measurements of HCV RNA per cell are transformed into a spatial point pattern (see <i><a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1003934#s4" target="_blank">Materials & Methods</a></i>). The amount of HCV RNA in those cells with the maximal HCV RNA content is subsequently reduced to the next lower level (red color) (Step I–Step II). For each of the different steps, spatial point patterns are produced, and a <i>Matérn cluster process</i> is fitted to each pattern to estimate . Step shows the last step before the cutoff criterion for the maximal cluster extension. For all subsequent steps of this example Pearson's chi-squared statistic for the point patterns indicated spatial heterogeneity for less than 95% of the bootstrap samples.