Predictive Modeling of Stochastic Transcriptional Burst Dynamics in Single Neurons via Hierarchical Transformer Networks with Adaptive Temporal Resolution

**Abstract:** Single-neuron transcriptional bursting is a fundamental process in neural circuit function, influencing synaptic plasticity, network activity, and ultimately, cognitive behavior. However, accurately modeling the complex spatio-temporal dynamics of these bursts remains a significant challenge. This paper introduces a novel predictive modeling framework leveraging hierarchical transformer networks augmented with adaptive temporal resolution control to capture the stochasticity and long-range dependencies inherent in neuronal transcription. Our approach combines aspects of time-series analysis, deep learning, and stochastic differential equations to provide a highly accurate and interpretable model, enabling both predictive simulations and insights into the underlying molecular mechanisms. This technology holds significant potential for drug discovery, personalized medicine, and understanding the pathogenesis of neurological disorders.