Substance identification by quantitative characterization of oscillatory activity in murine spinal cord networks on microelectrode arrays.

This paper presents a novel and comprehensive method to identify substances on the basis of electrical activity and is a substantial improvement for drug screening. The spontaneous activity of primary neuronal networks is influenced by neurotransmitters, ligands, and other substances in a similar fashion as known from in vivo pharmacology. However, quantitative methods for the identification of substances through their characteristic effects on network activity states have not yet been reported. We approached this problem by creating a database including native activity and five drug-induced oscillatory activity states from extracellular multisite recordings from microelectrode arrays. The response profiles consisted of 30 activity features derived from the temporal distribution of action potentials, integrated burst properties, calculated coefficients of variation, and features of Gabor fits to autocorrelograms. The different oscillatory states were induced by blocking neurotransmitter receptors for: (i) GABA(A); (ii) glycine; (iii) GABA(A) and glycine; (iv) all major synaptic types except AMPA, and (v) all major synapses except NMDA. To test the identification capability of the six substance-specific response profiles, five blind experiments were performed. The response features from the unknown substances were compared to the database using proximity measures using the normalized Euclidian distance to each activity state. This process created six identification coefficients where the smallest correctly identified the unknown substances. Such activity profiles are expected to become substance-specific 'finger prints' that classify unique responses to known and unknown substances. It is anticipated that this kind of approach will help to quantify pharmacological responses of networks used as biosensors.