Figure 16. MBON input and output similarity structure.

A comparison of MBON inputs and outputs through the lens of similarity structure. Left: cosine similarity of MBONs based on their effective PN inputs via KCs (computed as in Figure 15). Each cell of the heat map indicates the input similarity of the indicated pair of MBONs. A value of 0.5 indicates that the PN subpopulations conveying input to the two MBONs via KCs are half overlapping and half disjoint. Right: cosine similarity of MBONs based on their outputs to all neurons (unnamed neuronal fragments have been excluded). The resemblance between the left and right plots suggests that MBON outputs preserve some of the parallel structure in the PN-KC-MBON pathway. In each plot, the square in the diagonal is color-coded to indicate the MB lobe in which that MBON’s dendrites lie. The MBON names are color-coded to indicate their neurotransmitter, as indicated.

Figure 16—figure supplement 1. MBON output similarity by lobe.

Cosine similarity of MBONs based on the similarity of their outputs to all neurons (unnamed neuronal fragments have been excluded). Each cell of the heat map indicates the output similarity of the indicated pair of MBONs. This is the same data as plotted in Figure 16, except that the MBONs are organized by lobe, shown in the same order as in Figure 15. The preservation of lobe-wise sensory input structure in the MBONs suggests that MBONs may mediate multiple parallel pathways that transform sensory information into behavioral output. The MBON names are color-coded to indicate their neurotransmitter, as indicated.

Figure 16—figure supplement 2. Correlation of MBON output and PN input similarity.

Scatter plot of MBON output similarity vs. similarity of their inputs from PNs, quantifying the relationship between the structure of MBON outputs and their sensory inputs. Each point represents a pair of MBONs from different compartments. Output similarity is computed using the cosine similarity of outputs, as in Figure 16. PN input similarity is computed using cosine similarity of effective PN inputs, as in Figure 16. The highest 10 points on the y-axis correspond to the following pairs (fewer than 10 listed due to multiple MBONs of the same type): MBON31 (α′1a) and MBON32 (γ2); MBON26 (β′2d) and MBON31 (α′1a); MBON12 (γ2α'1) and MBON13 (α′2); MBON04 (β′2mp_bilateral) and MBON12 (γ2α′1); MBON26 (β′2d) and MBON32 (γ2); MBON07 (α1) and MBON14 (α3).

Figure 16—figure supplement 3. MBONs clustered by output similarity.

Cosine similarity between each MBON type and their target population. The ordering of MBONs was determined using average clustering. Unlike Figure 16 or Figure 16—figure supplement 1, the ordering of the MBONs in this figure does not take into account input structure or anatomical structure, allowing for unbiased grouping of MBONs based on their outputs. This reveals some clusters of MBONs with convergent outputs that span multiple lobes. Only named neurons were considered in the target population. MBON names are color-coded by neurotransmitter with predicted transmitters also underlined.

Figure 16—figure supplement 4. MBON morphologies grouped by output similarity.

Radial dendrogram shows that MBONs with similar downstream connectivity innervate different compartments. A dendrogram based on the MBON cosine similarity matrix (Figure 16—figure supplement 3) was extracted and the morphologies of the MBON in each group were plotted. Colored MBON names in each panel match the colors of the MBONs in the morphology images. MBONs can have similar downstream connectivity regardless of whether they innervate the same (MBONs 16, 17, and 28) or different (MBONs 23, 18, 06, and 02) compartments. The asterisk indicates MBONs (MBON10, 13, 15, 20, and 33) that could not be grouped with another partner; they are singletons where their downstream connectivity is different from all other MBONs.