A model study on the circuit mechanism underlying decision-making in Drosophila

✍ Scribed by Zhihua Wu; Aike Guo

• Publisher: Elsevier Science
• Year: 2011
• Tongue: English
• Weight: 837 KB
• Volume: 24
• Category: Article
• ISSN: 0893-6080
• DOI: 10.1016/j.neunet.2011.01.002

✦ Synopsis

Previous elegant experiments in a flight simulator showed that conditioned Drosophila is able to make a clear-cut decision to avoid potential danger. When confronted with conflicting visual cues, the relative saliency of two competing cues is found to be a sensory ruler for flies to judge which cue should be used for decision-making. Further genetic manipulations and immunohistological analysis revealed that the dopamine system and mushroom bodies are indispensable for such a clear-cut or nonlinear decision. The neural circuit mechanism, however, is far from being clear. In this paper, we adopt a computational modeling approach to investigate how different brain areas and the dopamine system work together to drive a fly to make a decision. By developing a systems-level neural network, a two-pathway circuit is proposed. Besides a direct pathway from a feature binding area to the motor center, another connects two areas via the mushroom body, a target of dopamine release. A raised dopamine level is hypothesized to be induced by complex choice tasks and to enhance lateral inhibition and steepen the units' response gain in the mushroom body. Simulations show that training helps to assign values to formerly neutral features. For a circuit model with a blocked mushroom body, the direct pathway passes all alternatives to the motor center without changing original values, giving rise to a simple choice characterized by a linear choice curve. With respect to an intact circuit, enhanced lateral inhibition dependent on dopamine critically promotes competition between alternatives, turning the linear- into nonlinear choice behavior. Results account well for experimental data, supporting the reasonableness of model working hypotheses. Several testable predictions are made for future studies.