Colocalization analysis

Small RNA colocalization analyses were carried out using Clus-DoC colocalization software, which is designed for single-molecule localization microscopy data (28). Five thousand images from each small RNA localization image were taken for analysis. Image alignment was done with fiducials in Zen software (Carl Zeiss Inc., Thornwood, NY, USA). As part of the ClusDoC co-clustering analysis, each pair of small RNA images were clustered using DBSCAN. A minimum cluster size of three points and an epsilon radius of 200 nm (with 40 nm histogram-smoothed contours) were manually tuned to produce the most accurate clustering of small RNA with DBSCAN in individual images. Colocalization analysis in ClusDoC was performed by calculating and comparing the density distributions of pairwise points between channels. For this analysis, a large maximum radius (Rmax) of 2500 nm and a step size of 10 nm were selected for the calculation of the discrete distributions due to the density of the dataset.