Code for simulating and analyzing a dynamic model of the mushroom body, including olfactory receptor neurons (ORNs), projection neurons (PNs), GABAergic local neurons (LNs) of the antennal lobe, Kenyon cells (KCs) and the GABAergic anterior paired lateral neuron (APL).

For more information about the model, please see the accompanying bioRxiv paper.

This model simulates Kenyon cell responses to each of 110 odors from the Hallem & Carlson ORN odor response dataset.

To run the model with default settings and get KC activations to one or more odors, add the mushroomBody repository to your Matlab search path and call run_rate_model(odorList) where odorList is a vector of indices between 1 and 110, corresponding to the odors you want to simulate in the Hallem & Carlson dataset. So run_rate_model(1:110) would simulate responses to all odors; run_rate_model(94) would only simulate responses to odor 94, which is isopentyl acetate.

To load/inspect the Hallem & Carlson dataset itself (with default settings), call ORN = buildORNs({}). The names of all odors are listed in ORN.odornames.

run_rate_model returns three variables:

• KC_spikes, an N (# neurons) by S (# odor stimuli) matrix whose entries reflect the number of spikes fired by each Kenyon cell in response to each odor in odorList.
• fr_active, an S-length vector of the number of Kenyon cells that responded to each odor.
• odorspercell, an N-length vector of the number of odors each Kenyon cell responded to.

This repository contains code to reproduce all model variations presented in the paper. Use the script run_rate_model_wrapper to simulate multiple runs of the mushroom body model; parameters that can be varied are explained in get_MB_default_settings.

• analyses_from_paper contains code for generating the plots in the paper.
• constructing_model contains code to initialize model ORNs, PNs, LNs, and KCs.
• learning contains scripts to experiment with learning and odor generalization by either a Perceptron or a biological learning rule.
• simulation_helpers are scripts called in the process of simulating Projection neuron and Kenyon cell dynamics.
• util are plotting/data-searching helper functions

If you use or modify this model, please cite the accompanying paper:

Kennedy, A. (2019). Learning with naturalistic odor representations in a dynamic model of the Drosophila olfactory system. bioRxiv.