Emergence of biological responses tuned to predictive codes in Drosophila larvae

Authors: Antony W. N’dri, Frédéric Guyard

Presentation type: Poster at SNUFA 2026 online workshop (4-5 Nov 2026)

Abstract

Recent advances in brain imaging has allowed for the development of full connectomes of insects’ brains such as Drosophila melanogaster. The limited number of neurons of Drosophilae make it a promising candidate to uncover the elementary functions and properties behind biological sensory and decision-making neural responses. Here, we study a simple decision-making circuit found in Drosophila larvae. This circuit interestingly promotes a sequence of defensive actions (hunch-bend) in the larva after stimulation with a mechanosensory stimulus. Using the e-prop algorithm, we first train biologically-constrained spiking networks to exhibit such biological responses on Poisson spiking inputs as well as the N-MNIST digits and demonstrate that these responses are tuned to compact neural codes (that have few spikes and exhibit high sparsity). We apply this approach to a simple anomaly detection task as an application. Our results show that such a Drosophila larvae neural circuit can best signal deviations to usual sensory inputs when they are encoded with a predictive coding approach to have few spikes, while networks tuned to larger amounts of spikes are less sensitive to such input changes. Interestingly, this suggests that even the neural circuits of small organisms such as Drosophila melanogaster might have evolved to favor energy-efficient codes as prescribed by the predictive coding theory.

Are you the author and do you want to add some related links to this page? If so, check out our guide to adding related links and videos.