Fig. 4: Integration of ACh dynamics improves the odor identity representation by neuronal activity.

a Map of decoding-accuracy gain averaged across flies. b Average accuracy gain in each brain region. c Voxel-level multiple-brain-region odor identity classification accuracies. d, e Low-dimensional manifolds by G7f (d) and both channels (e) of the same fly as in Fig. 3d. Arrows are in arbitrary units but consistent across dimensions. f–h Metrics of the manifolds. f Distance to the origin. g Inter-class distance. h Intra-class distance. i Functional connectivity matrices (top) and mappings of communities and connectivity change in the physical space (bottom) of the voxels within the MBML. Three columns are the functional connectivity of G7f, rACh, and the difference in the deflation ratio between them. Connectivity complementation (top, red rectangles) and emphasis (top, black rectangles) to G7f by rACh are shown. Complementation is reflected in the different distribution of functional connections (bottom, red rectangles). Community divisions are similar for the two channels (bottom), and voxels with the greatest increased connectivity gather in space (bottom, coral). j Similar to (i), but for the LH. k Difference in increased deflation ratios within and between clusters of the functional connectivity of rACh compared to G7f. l Average clustering coefficient for rACh. m Average clustering coefficient difference between two channels. n Node centrality distribution. o Distance of the node centrality distributions between two channels. In (k–m), and (o), brain regions on the left side of the dashed line are regions with accuracy gain, and hypothesis testing is performed for them toward the average level of the LH, EPA, AOTU, and IVLP. n = 10 flies co-labeled by G7f and rACh, mean ± s.e.m. Each light-colored line represents a fly in (c), (f–h), and (k–o). One-sided Wilcoxon signed-rank test in (b), (c), (k–m), and (o); Two-sided Wilcoxon signed-rank test in (f–h); Benjamini/Hochberg multi-comparison correction in (f–h). Source data are provided as a Source data file. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, ns - not significant (P > 0.05, not shown in b, k–m, o). Scale bar: 50 μm.