PREreview del Odour representations supporting ethology-relevant categorisation and discrimination in the Drosophila mushroom body

In this study, the authors investigate how specific wiring patterns between projection neurons (PNs) and Kenyon cells (KCs) influence olfactory processing, a question with broad significance as it reveals the general principles by which neural circuits convert sensory inputs into behaviorally relevant codes. Although PN–KC connectivity has long been assumed to be random, recent detailed connectome analyses reveal clear deviations from randomness as suggested by preferential co-arborization of certain PN types and biased connections between PNs and KCs. However, the functional significance of these newly identified deviations has remained unclear. The authors propose that deviations from random wiring enhance ethological categorization, a conclusion strongly supported by convergent evidence from connectomes, olfactory responses of KCs (new experimental data from the current study), and computational modeling. This work reconciles the classic hypothesis that random connectivity supports odor discrimination with recent discoveries of structured PN–KC connections. It provides a mechanistic explanation for how dissimilar odors can be recognized as categories, an aspect of olfactory coding that has been largely unknown. Overall, this study greatly advances our understanding of odor representation in the fly olfactory system. Below are minor comments that may help further strengthen the manuscript:

(1)     In figure 1(d) and similar plots in supplementary figures, it would be helpful for readers to also see the absolute numbers in addition to percentages, given that KC subtype abundances vary (as shown in supplementary data figure 1(c)).

(2)     The random non-uniform connectivity schematic in Figure 1(e) may give a misleading first impression of altered spatial distribution because most boutons appear on the right side. It would be clearer to emphasize the counts rather than the spatial layout and explicitly show on the schematic that the bouton numbers are proportional to the fractions observed in the hemibrain dataset.

(3)     Although the authors perform activity imaging on danger-related odors, the connectome analysis includes comparatively little information about danger-responsive PNs (DPNs), in contrast to FPNs and RPNs. It would be informative to also compare the wiring patterns of DPNs with FPNs/RPNs, and explore whether their connectivity differences relate to the physiological properties observed.

(4)     It would be helpful to add a brief explanation, either in the supplementary figure legends or in the Methods, about why Euclidean distances are used in Figure 3(b) whereas cosine distances are used in Supplementary Figures 3–6, and why these methods lead to the different results. This clarification would help readers more fully appreciate the conclusions presented in the main figure.

Competing interests

The authors declare that they have no competing interests.

Use of Artificial Intelligence (AI)

The authors declare that they did not use generative AI to come up with new ideas for their review.