Computing in neuronal networks with plasticity via all-optical bidirectional interfacing

All-optical techniques play a pivotal role in the direct interrogation of biological neural networks and the investigation of their computational properties. Such techniques combine spatiotemporally precise optogenetic stimulation at cellular resolution to manipulate neuronal states in the network with optical imaging of fluorescent genetically-encoded indicators for simultaneous readout of neuronal activity [1, 2]. The biophysical models of various optogenetic tools have been studied computationally in single neurons, accurately fitting experimental data [3, 4]. However, the effects of optogenetic manipulations with complex excitation and inhibition patterns have been explored in less detail [5, 6], especially in the context of the plasticity induction and the computation in neuronal networks [7]. Thus, we can ask the following questions: (i) how can different stimulation protocols modify functional connectivity in isolated neuronal systems, and (ii) what computations can be practically realized with such an approach?
In this work, we introduce optogenetic manipulation and fluorescence readout in spiking neuronal networks with recurrent connectivity and spike-timing dependent plasticity. We propose a protocol for "programming" the functional connectivity of the network and use this protocol to implement a simple computational task, memory storage and recall. Notably, during the "programming" phase, the protocol separates interdependence between the activity of the network and the induction of the associated changes in functional connectivity. It also permits a simplification of the description of the neuronal network as a dynamical system and guarantees reproducible initial conditions for its temporal evolution. This work represents a significant advance by providing a realistic simulation of a neuronal system with a bidirectional optogenetic interface that takes into account both biophysical and computational aspects and suggests a strategy for further investigation of computation in neuronal networks.