Affinity propagation clustering of Drosophila melanogaster neurons

Supporting information for Costa et al. (2014) NBLAST: Rapid, sensitive comparison of neuronal structure and construction of neuron family databases.

We have used affinity propagation clustering of our NBLAST scores to cluster the full FlyCircuit dataset of 16,129 neurons into 1,052 distinct clusters of highly related neurons. These clusters have been organised into superclusters, enabling both exploration of the dataset and the matching of individual clusters to morphological types reported in the literature, with these webpages being provided as a tool to enable this exploration. Each supercluster is detailed on its own page, linked to by the images below. These pages provide an interactive 3D model of each supercluster with links to pages detailing the clusters contained within. Interactive 3D models of each individual cluster are provided on these pages, as well as information on the constituent neurons and similar clusters as determined by NBLAST.


Raw data were provided by as described in Three-Dimensional Reconstruction of Brain-wide Wiring Networks in Drosophila at Single-Cell Resolution by Ann-Shyn Chiang et al. These were then registered into a common template space and downsampled by a factor of 2 in x and y, binarized with a threshold of 1 and then skeletonized using the Fiji plugin 'Skeletonize (2D/3D)'. Dot properties for each neuron skeleton were extracted following the method we have previously described in Masse et al. (2012), using the dotprops function of our new nat package for R.


3D visualisation and analysis was performed using our open-source R package, nat. NBLAST is implemented in a further open-source R package, nat.nblast. Analysis code specific to the FlyCircuit dataset is available in a dedicated R package, flycircuit, with a package vignette showcasing the main tools that we have developed.