Figure 3. Neuroanatomical models with nat.

(a) We accessed the InsectBrainDB.org via insectbrainr to obtain template brains for different species of insect (Brandt et al., 2005; de Vries et al., 2017; El Jundi et al., 2018; Heinze and Reppert (2012); Kurylas et al. (2008); Løfaldli et al. (2010); Stone et al. (2017); Zhao et al., 2014). The package insectbrainr converts retrieved OBJ files into hxsurf objects, which contain one set of 3D points for each whole brain, and then different sets of edges between these points to form 3D neuropil subvolumes. These subvolumes were already defined by expert annotators. Their volume are compared across insect brain, normalised by total brain size. Insect template brain data curated by: S. Heinze, M. Younger, J. Rybak, G. Pfuhl, B. Berg, B. el Jundi, J. Groothuis and U. Homberg. (b) We can create our own subvolumes by pulling synaptic neuron reconstructions (Berck et al., 2016) from a first-instar larva EM dataset (Ohyama et al., 2015) (a public CATMAID instance hosted by Virtual Fly Brain), extracting dendritic post synapses from olfactory projections neurons, and using synapse clouds from neurons of the same cell type, to define glomerular volumes by creating a bounding volume, i.e an α-shape or convex hull. Their volumes can then be calculated, and correlated with the number of presynapses the same neurons make in two higher-order brain regions, the lateral horn and the mushroom body calyx. (c) Volumes can be used to analyse skeleton data. In (c) we look again at olfactory projection neurons, this time from an adult fly EM dataset (Zheng et al., 2018) and use the nat function pointsinside with standard neuropil volumes (Ito et al., 2014) to find the numbers of presynapses GABAergic and cholinergic olfactory projection neurons from the antennal lobe make in different neuropils. These neuropils exist as a hxsurf object in our R package nat.flybrains.

Figure 3—figure supplement 1. Superxovel analysis with nat.

(a) The basic connectivity scheme of the lateral horn, a second order olfactory centre in insects. Olfactory projection neurons (in orange) connect to lateral horn neurons (in cyan) (Frechter et al., 2019). (b) To create anatomically meaningful continuous voxels for the lateral horn, rather than random contiguous partitions of our standard neuropil space (Ito et al., 2014), we first removed the highest Strahler order branch (assign_strahler) from projection neuron axons’ so that their sub-branches could be clustered into 25 separate groups (nblast). For each cluster, a 3-D weighted kernel density estimate was generated based on 3D points (xyzmatrix) extracted from clustered sub-branches, using the R package ks (Duong, 2007). Points were spaced on neurites at 1 μm intervals (resample) and weighted as 1/total number of points in the cluster, so that supervoxels could be directly compared. (c) An 'inclusion' score for each neuron was calculated for each supervoxel by summing the density estimate for each point in the chosen arbor, again sampled at 1 μm intervals, and normalised by the total number of points in each arbor. An atlas of the lateral horn, colouring supervoxels by the modality/valence of their strongest input neurons. (d) A ‘projection’ was calculated between each lateral horn voxel and each lateral horn output voxel based on the number of neuronal cell types that have processes in both and the density of this arborisation. An atlas of the lateral horn output regions can then be made, colouring supervoxels by the modality/valence of their strongest input lateral horn supervoxels.