Veri Tabanlı Hippokampal CA1 Devresinin Basitleştirilmesi

Biophysically realistic neuron models provide emergent neural and circuit behaviors due to their data-driven approach and high sensitivity in representing the ionic currents. These models comprise multiple compartments for representing dendrites, which allows getting high accuracy during dendrosomatic and somatodendritic signals. An average pyramidal cell model having around 170 compartments and 1000 synapses makes up a significant cost in computational time and memory load. When a high number of cells are considered, an ordinary computer becomes unable to simulate such circuits. With the recent developments in neuron simplification algorithms, the voltage deflection timings at the soma can be captured by preserving the transfer impedance of excitatory \& inhibitory synapses to soma. It has been shown that the simplified cortical neuron models can get a high spike-synchrony with up to 250x faster simulation time. Despite its performance at single cell level, the application of neuron reduce algorithm to circuit level analysis has not been shown yet. This work extends the previous work by applying it to various hippocampal morphologies and shows its performance in single cell and network level simulations. The validations include total neuron activation rates by morphology, inter-spike intervals of the circuits, raster plots and average firing rates throughout the circuit in experiments including somatic current injection, miniature post synaptic potentials, external innervation from Schaffer Collaterals and LFP simulations. We show the successfully replicated emergent behaviors and the limitations of the simplification algorithm on neurons firing in broad frequency regimes and further present optimization techniques for sub-optimal neurons.