Towards tracking homeostais on high-density multi-electrode arrays

Homeostatic plasticity is one of the key mechanisms ensuring the remarkable adaptive abilities of the brain. However, this is still a relatively scantly explored branch of both experimental and computational neuroscience - in particular on a large, multi-neuronal scale. With recent advance in recording techniques, the lack of experimental data can be easily overcome – novel multielectrode arrays allow for high-density recordings from in vitro cultures consisting of thousands of neurons. What is needed to complement this rich data is analysis techniques that would be able to shed some light on the mechanism of the underlying process – in contrast to most conventional analysis techniques, such as firing rates, correlations or inter-burst intervals, which provide little more than descriptive information. In search for measures able to capture more complex phenomena, over the last decade a new approach has been developed - pairwise maximum entropy modelling (MaxEnt). It is a statistical model that fits two sets of parameters to explain the probability of spiking patterns in the network: individual neuron parameters that could be interpreted as excitability; and pairwise interaction parameters that could be interpreted as the functional connection strength between neurons. Successful application of this model to a variety of recordings has helped reevaluate the importance of neuronal interactions in shaping network activity (Schneidman et al., 2006; Shlens et al., 2006). Additionally, the shortcomings of MaxEnt in certain cases can serve as an indicator of higher-order interactions between neurons (Ohiorhenuan et al., 2010). In present work we examine the extent to which the statistics of MaxEnd model fits and parameters can assist in understanding different modes of activity of a neuronal culture – specifically, along the duration of a homeostatic experiment. Neural activity from primary neuron cultures was recorded with the 4096 channel Active Pixel Sensor (APS) MEA, allowing for reliable isolation of single unit activity at near-cellular resolution (Berdondini et al., 2009). 20-minute datasets were obtained at different stages of homeostatic compensation during and after long-term CNQX application. For the data sets with a stationary activity state, large numbers of four-unit MaxEnt models were constructed for randomly chosen neurons on two spatial scales. Comparison of the statistics of the fits and parameters across the scales and across conditions indicates that different activity modes exhibit different profiles of local clustering and higher-order interactions.