Spatiotemporal Neural Activity Tracking and Continuous Transcriptional Variation of Neurons
Identifying and tracking cell location in long-term longitudinal studies is critical for identifying large-scale time-dependent neuronal activity variations. Population cell tracking across multiple sessions is complicated by non-rigid transformations induced by noise, cell movement and imaging field shifts. In this text we develop SCOUT (Single-Cell SpatiOtemporal LongitUdinal Tracking), a fast, robust cell tracking method utilizing multiple cell-cell similarity metrics, probabilistic inference, and an adaptive clustering methodology to perform cell identification across multiple calcium imaging sessions. We then apply SCOUT to study variations of firing activity coincident with contextual discrimination and neural circuit negation. Next, we investigate the relationship between firing activity and transcriptomics in a single cell type, showing that transcriptional gradients can be associated to subtle variations in neuronal firing activity, which then motivates the development of scGradient, a machine learning algorithm for identifying continuous transcriptional gradients across and within cell types.