bioRxiv2020
Microelectrode arrays (MEAs) are valuable tools for electrophysiological analysis at a cellular population level, providing assessment of neural network health and development. Analysis can be complex, however, requiring intensive processing of large high-dimensional data sets consisting of many activity parameters. As a result, valuable information is lost, as studies subjectively report relatively few metrics in the interest of simplicity and clarity. From a screening perspective, many groups report simple overall activity; we are more interested in culture health and changes in network connectivity that may not be evident from basic activity parameters. For example, general changes in overall firing rate - the most commonly reported parameter - provide no information on network development or burst character, which could change independently. Our goal was to develop a fast objective process to capture most, if not all, the valuable information gained when using MEAs in neural development and toxicity studies. We implemented principal component analysis (PCA) to reduce the high dimensionality of MEA data. Upon analysis, we found that the first principal component was strongly correlated to time, representing neural culture development; therefore, factor loadings were used to create a single index score - named neural activity score (NAS) - reflective of neural maturation. To validate this score, we applied it to studies analyzing various treatments. In all cases, NAS accurately recapitulated expected results, suggesting this method is viable. This approach may be improved with larger training data sets and can be shared with other researchers using MEAs to analyze complicated treatment effects and multicellular interactions. Author SummaryAnalyzing neural activity has important applications such as basic neuroscience research, understanding neurological diseases, drug development, and toxicity screening. Technology for recording neural activity continues to develop, producing large data sets that provide complex information about neuronal function. One specific technology, microelectrode arrays (MEAs), has recently given researchers the ability to record developing neural networks with potential to provide valuable insight into developmental processes and pathological conditions. However, the complex data generated by these systems can be challenging to analyze objectively and quantitatively, hindering the potential of MEAs, especially for high-throughput approaches, such as drug development and toxicity screening, which require quick, simple, and accurate quantification. Therefore, we have developed an index for simple quantification and evaluation of neural network maturation and the effects of perturbation. We present validation of our approach using several treatments and culture conditions, as well as a meta-analysis of toxicological screening data to compare our approach to current methods. In addition to providing a simple quantification method for neural network activity in various conditions, our method provides potential for improved results interpretation in toxicity screening and drug development.