Advanced Data Analysis methods to explore the neuronal dynamics in brain-on-a-chip models

Objective In the field of neuroengineering, the development of reliable in vitro neuronal networks is particularly significant, especially when supported by data analysis methods capable of exploring and dissecting the diverse dynamical regimes that may characterize neuronal assemblies under different conditions (e.g., spontaneous vs evoked/modulated activity). The aim of my PhD project was to develop and apply advanced methods specifically designed for electrophysiological analyses of in vitro models, enabling a deeper understanding of neuronal dynamics. In particular, I focused on three main topics: the implementation of user-friendly software for dose-response curve fitting, the creation of a custom detection method for the identification of electrophysiological signals from neurospheroids coupled to Micro-Electrode Arrays (MEAs), and the development of a computational model exploiting Evolutionary Game Theory to simulate and analyse neuronal communication. Each issue aimed at extracting quantitative metrics that describe the electrophysiological behavior of experimental neuronal models, providing tools to get insights into experimental outcomes. Approach The first goal involved the design of a software to generate dose-response curves and determine the drug concentrations at which a relevant effect on a specific metric (e.g., the spiking frequency) of in vitro neuronal cultures occurs. This software was tested on hippocampal and cortical networks, revealing differences between homogeneous (composed of a single neuronal type) and heterogeneous (with interacting neuronal populations) models. The second topic required the development of a novel detection method tailored to the signals generated by neurospheroids. Detection methods employed for planar networks proved ineffective due to the summation of neuronal contributions in the spheroids, producing oscillatory signals. By employing wavelet analysis, I developed a method that explores both the temporal and frequency domain, through the computation of a spectrogram of the electrophysiological signal. Finally, I developed a computational model using Hindmarsh-Rose equations to simulate individual neuron dynamics, while the connectivity between neurons was modeled exploiting principles from Evolutionary Game Theory. This model aims to replicate the neuronal activity observed in vitro and explore how different game-theory-derived parameters impact the emergent dynamics. Main results The dose-response curve software effectively demonstrated the importance of using heterogeneous neuronal models for drug testing, as the extracted IC50 values varied significantly between homogeneous (i.e., only one neuronal type) and heterogeneous networks (at least two different neuronal types). The detection method for neurospheroids, designed to handle the complex, oscillatory signals of densely packed neuronal clusters, was able to identify electrophysiological activity in both time and frequency domains. This method was validated on a dataset comprising both neurospheroids and neuroassembloids (coupled spheroids), where it detected significant differences in the percentage of frequency band expression, depending on the neuronal type and the characteristics of the spheroids and the assembloids. The Evolutionary Game Theory-based model proved to be not only a mathematical tool for generating electrophysiological patterns but also a computational approach capable of deriving game theory strategies present within a neuronal network. This feature highlights the potential of the model to classify network characteristics, such as distinguishing between cortical and hippocampal neuronal types, or between modular, heterogeneous, and three-dimensional networks, through the evaluation of the extracted Evolutionary Game Theory parameters. Significance The developed computational tools have broad implications for the analysis of electrophysiological data of in vitro neuronal models. The dose-response curve software aims at being used in preclinical studies. The detection method for neurospheroids enhances the study of three-dimensional neuronal models, which better mimic in vivo brain conditions due to their neuronal heterogeneity and structural complexity. The Evolutionary Game-Theory-based computational model holds the potential to be a powerful tool for classifying neuronal network types and understanding how connectivity and competition within neuronal populations shape emergent dynamics, providing insights into both healthy and pathological brain function.