Figure 4. Cell typing with nat.

(a) Neurons from a dense reconstruction from EM data of the mouse retina inner plexiform layer (Helmstaedter et al., 2013) can either be NBLAST-ed in situ (upper) or after alignment by their principal axes in 3D space (lower) in order to make a first pass at defining, finding or discovering morphological neuronal cell types using NBLAST. (b) A tSNE plot visualising the results of an aligned NBLAST of neurons in A, coloured by the manually annotated cells types seen in Figure 2c, with shapes indicating the cell class. (c) Manual sorting using the interactive nat functions nlscan and find.soma or find.neuron can help identify misclassifications and make assignments.

Figure 4—figure supplement 1. Cell typing zebrafish neurons with nat.

(a) Light-level neurons registered to a standard brain for the larval zebrafish (Kunst et al., 2019) can be read into R using fishatlas, and all transformed onto the right hemisphere using the mirroring registration generated by Kunst et al. (b) They can then be NBLAST-ed to discover new cell types. 25 NBLAST clusters are shown, generic exemplary code shown in Monaco font. (c) Upper, shows this process for a small group of neurons from the telencephalon. Mirroring neurons to the same side produces cohesive clusters. Lower, neurons can be cut up (pruned) relative to user defined coordinates to, in this case, divide them by telencephalon region. NBLAST can be run separately on these two groups, and the results were compared using a tanglegram to see whether neurons that cluster together in the posterior telencephalon, also cluster together in the ventral telencephalon.

Figure 4—figure supplement 2. Cell typing vinegar fly neurons with nat.

The same basic process can be done for FlyCircuit neurons (Chiang et al., 2011), here subsetted (in_volume) to those neurons that have arbor in both the antennal lobe and lateral horn (i.e. olfactory projection neurons).