Publication
Long-range projections are the structural basis for inter-areal interactions across a cortical hierarchy. These connections are classified based on their laminar termination profiles into feed-forward, when they connect a lower-order to a higher-order area in the hierarchy, or feedback in the opposite direction. Top-down influences on sensory perception, such as attention, expectation or perceptual task, are thought to be communicated through feedback projections. It is not well understood, however, what are the mechanisms behind these influences or the exact nature of the information carried by feedback pathways. This thesis presents the first biophysically-detailed simulation-based approach to the study of interactions between cortical areas and their role in sensory processing. This model is based on a large-scale, biophysically-detailed model of rodent sensory cortex, developed as a collaborative effort within the Blue Brain Project. In terms of anatomical aspects, the base model includes: a novel atlas-based approach to placing reconstructed neuronal morphologies inside the cortical volume, a realistic model of thalamocortical innervation, and data-driven algorithms for the generation of local and long-range connections. Notably, the definition of cortico-cortical projections requires a flatmap, a two-dimensional representation of the horizontal extents of the cortex, to establish a topographical mapping between areas; I present here an algorithm to generate flatmaps from digital brain atlases. Among physiological aspects, the base model features: diverse single-cell electrical behaviors, rich parameterized synaptic physiology and in vivo-like spontaneous and evoked activity. The two-area model shares most characteristics with the base model, but provides a reduced setting in which to study inter-areal interactions. Each of the two model areas (X and Y) consists in over 200,000 neurons of 60 morphological types distributed across six layers. Area X represents a primary sensory area, receiving thalamocortical inputs and sending feed-forward outputs to area Y, while area Y represents a higher-order area receiving inputs only from area X and sending feedback outputs to area X. Both areas additionally receive background noisy inputs. The model exhibits a clear cortico-cortical loop, wherein responses to sensory inputs in area X have two components in time. The early component is stimulus-driven and consists in the activation of stimulus-specific functional cell assemblies. The late component is driven by feedback from area Y and has no stimulus specificity. Early responses in area Y have partial stimulus specificity in a way consistent with the topography of long-range connections. The model predicts a prominent role of feed-forward pathways originating in layer 5 and feedback pathways originating in layer 4 in producing the observed responses. When a stimulus arrives simultaneously with the feedback-driven component, an approximate linear super