Fig. 2.

Alignment of ORNs $\to$ LN synaptic count vectors with odor representations in ORN activity. (A) Ca²⁺$\mathrm{\Delta }F/F_0$ activity patterns ${\{{\mathbf{x}}^{(t)}\}}_{\text{data}}$ in ORN somas in response to 34 odors (separated by vertical gray lines) at 5 dilutions (${10}^{-8},...,{10}^{-4}$) from ref. 5. See SI Appendix, Fig. S3 for odor labels and scaled ${\{{\mathbf{x}}^{(t)}\}}_{\text{data}}$. (B) Correlation between the four ORNs $\to$ LN${}_{\text{type}}$ synaptic count vectors ($w_{\text{LNtype}}$ for BT, BD, KS, and P0) with the odor representations ${\{{\mathbf{x}}^{(t)}\}}_{\text{data}}$ from (A). Slash: significant at 0.05 level; cross: significant at 0.05 FDR (false discovery rate) (31). P-values calculated by shuffling the entries of each ${\mathbf{w}}_{\text{LNtype}}$ (50,000 permutations). (SI Appendix, Figs. S4A and S5). (C) ORNs $\to$ Broad Trio synaptic count vector ${\mathbf{w}}_{\text{BT}}$ superimposed with ORN activity patterns ${\mathbf{x}}^{\text{(A)}}$ and ${\mathbf{x}}^{\text{(B)}}$ in response to the ligands 2-heptanone (odor A) and 2-acetylpyridine (odor B) at dilution ${10}^{-4}$. y-axis: ORN, follows order of (A). (D) Scatter plot representation of (C). ${\mathbf{w}}_{\text{BT}}$ is more strongly tuned to ${\mathbf{x}}^{\text{(A)}}$ ($r=0.6$, $P=0.004$) than to ${\mathbf{x}}^{\text{(B)}}$ ($r=0.14$, $P=0.3$). P-values not adjusted for multiple comparisons. (E) ${\mathbf{w}}_{\text{BT}}$ superimposed on the 1^st PCA direction of ${\{{\mathbf{x}}^{(t)}\}}_{\text{data}}$. $y$-axis: ORN, follows order of (A). (F) Scatter plot representation of (E) ($r=0.65$, $P=0.001$). P-values are not adjusted for multiple comparisons. (G) LN “connectivity tuning curves”: correlation coefficients sorted in decreasing order from (B) for each ${\mathbf{w}}_{\text{LNtype}}$. (H) Correlation coefficient r between the top 5 PCA directions of ${\{{\mathbf{x}}^{(t)}\}}_{\text{data}}$ and the four ${\mathbf{w}}_{\text{LNtype}}$s (SI Appendix, Fig. S6 A, B, and E). Two-sided P-values calculated by shuffling the entries of each ${\mathbf{w}}_{\text{LNtype}}$ (50,000 permutations). *: significance at 0.05 FDR.