StormGraph: A graph-based clustering algorithm for the analysis of super-resolution microscopy data

With super-resolution microscopy techniques such as Direct Stochastic Optical Reconstruction Microscopy (dSTORM), it is possible to image fluorescently labeled proteins on a cell membrane with high precision. Often, the extent to which such proteins cluster is biologically meaningful; for example, in B-cells, clustering of the B-cell receptor (BCR) is associated with increased intracellular signaling and B-cell activation, and spontaneous BCR clustering can cause chronic active BCR signaling that results in an aggressive B-cell malignancy. Computational methods are therefore needed to make quantifiable comparisons between the observed clustering in different data sets, such as for different cell types or different experimental conditions. Inspired by the success of graph-based clustering algorithms such as PhenoGraph in other research areas, we developed StormGraph, a graph-based clustering algorithm for analyzing Single Molecule Localization Microscopy (SMLM) data such as would be obtained by dSTORM. This talk will present StormGraph, which distinguishes clusters from random background and assigns individual localizations to specific clusters, allowing for a detailed analysis of statistics such as cluster area. The utility of StormGraph will be illustrated on dSTORM data of BCRs imaged on malignant B-cells.