Modelling structure-function relationships from neurons to networks in the cerebral cortex

Abstract

Using a multi-scale model based on extensive experimental data, I explored structure-function relationships in cortex. At a macro-scale, I used ANNs with cortical features to show that sparse connectivity enables efficient information processing in large and recurrent networks, as are found in cortex. I propose that cortex-inspired ANNs may be a useful tool to disentangle the relevance of other properties of cortical connectivity and neuronal function for information processing in a network. Incorporating more biological detail on the neuron and network level, we used a model reduction approach to explain the computation performed by L5PTs following a sensory stimulus. We showed that the input-output computation is largely preserved across morphologically and biophysically diverse neurons, and found that the main source of variability in sensory responses and receptive field shapes is differences in synaptic input. Finally, we investigated the relevance of subcellular scale wiring specificity in the TC-L5PT pathway. We found that this wiring specificity, combined with intracortical sensory-evoked excitation and in certain circumstances prestimulus input, is necessary for bursts observed in sensory-evoked responses which may ignite conscious sensory processing.