Specific combinations of volatile chemicals are detected by the nose and are interpreted by the brain as signatures for the presence of relevant objects. In natural environments, chemicals from many objects mix in the air prior to reaching the nose, making the task of identifying the underlying objects much more difficult. In such rich environments, mechanisms of attention are used by sensory systems to prioritize perception of specific objects over others. These mechanisms are believed to rely on feedback connectivity from central to peripheral sensory brain regions. Such feedback connectivity is abundant in the olfactory system, but how it serves to prioritize odors is currently unknown.
We built a biologically-realistic model of the olfactory bulb and piriform cortex to study the mechanisms of odor-specific attention. The model captures key properties of feedforward and feedback circuitry and takes into account the nonlinear interactions between odorants at the olfactory epithelium. A contextual input that defines the target odor is fed into piriform cortex. This input modifies basal activity within piriform cortex. Via feedback piriform modifies olfactory bulb dynamics. We modelled naturalistic mixtures to test the effectiveness of different mappings of feedback connectivity and show that a mapping in which cortical neurons avoid innervating glomeruli they receive inputs from, but otherwise project randomly to the olfactory bulb, is efficient in prioritizing the target odor. Moreover, we demonstrate that the modified input impacts each neuron in olfactory cortex almost insignificantly, while the ensemble representation of a prioritized odor increased dramatically. Our model explains the relationship between feedback projections and attention, provides predictions for feedback connectivity, and demonstrates the unique contribution of collective dynamics for supporting behavioral functions.