The map of synaptic connectivity among neurons in the brain shapes the computations that neural circuits may perform. Inferring the design principles of neural connectomes is, therefore, fundamental for understanding brain development and architecture, neural computations, learning, and behavior. We learn probabilistic generative models for the connectomes of the olfactory bulb of zebrafish, part of the mouse visual cortex, and of C. elegans. We show that in all cases, models that rely on a surprisingly small number of simple biological and physical features accurately predict the existence of individual synapses and their strength, distributions of synaptic indegree and outdegree of the neurons, frequency of sub-network motifs, and more. Furthermore, we simulate synthetic circuits generated by our model and show that they replicate the computation that the real circuit performs. Thus, our results reflect surprisingly simple design principles of real connectomes. We then explore the architectural features that may shape the computation that connectomes may carry. We measure the similarity of simulated spiking neural networks of neurons in terms of their response to different stimuli, and learn a functional metric between networks based on their synaptic differences. We show that unlike common graph theory tools, our metric accurately predicts the similarity of novel networks, relying on a sparse set of architectural features. We then identify potential key architectural features that control the computations that particular connectomes may implement.

The human brain displays a rich repertoire of states that emerge from the microscopic interactions of cortical and subcortical neurons. Unfortunately, difficulties inherent within large-scale simultaneous neuronal recording limit our ability to link biophysical processes at the microscale to emergent macroscopic brain states. Here we introduce a microscale biophysical network model of layer-5 pyramidal neurons that display graded coarse-sampled dynamics matching those observed in macroscale electrophysiological recordings from macaques and humans. We invert our model to identify the neuronal spike and burst dynamics that differentiate unconscious, dreaming, and awake arousal states and provide novel insights into their functional signatures. We further show that neuromodulatory arousal can mediate different modes of neuronal dynamics around a low-dimensional energy landscape, which in turn changes the response of the model to external stimuli. Our results highlight the promise of multiscale modelling to bridge theories of consciousness across spatiotemporal scales.