Predicting single-cell transcription dynamics even when the central limit theorem fails

Abstract

Despite substantial experimental and computational efforts, mechanistic modeling remains more predictive in engineering than in systems biology. The reason for this discrepancy is not fully understood. Although randomness and complexity of biological systems play roles in this concern, we hypothesize that significant and overlooked challenges arise due to specific features of single-molecule events that control crucial biological responses. Here we show that modern statistical tools to disentangle complexity and stochasticity, which assume normally distributed fluctuations or enormous datasets, don't apply to the discrete, positive, and non-symmetric distributions that characterize spatiotemporal mRNA fluctuations in single-cells. We demonstrate an alternate approach that fully captures discrete, non-normal effects within finite datasets. As an example, we integrate single-molecule measurements and these advanced computational analyses to explore Mitogen Activated Protein Kinase induction of multiple stress response genes. We discover and validate quantitatively precise, reproducible, and predictive understanding of diverse transcription regulation mechanisms, including gene activation, polymerase initiation, elongation, mRNA accumulation, spatial transport, and degradation. Our model-data integration approach extends to any discrete dynamic process with rare events and realistically limited data.\n\nSignificance StatementSystems biology seeks to combine experiments with computation to predict complex biological behaviors. However, despite tremendous data and knowledge, most biological models make terrible predictions. By analyzing single-cell-single-molecule measurements of mRNA in yeast during stress response, we explore how prediction accuracy is controlled by experimental distributions shapes. We find that asymmetric data distributions, which arise in measurements of positive quantities, can cause standard modeling approaches to yield excellent fits but make meaningless predictions. We demonstrate advanced computational tools that solve this dilemma and achieve predictive understanding of many spatiotemporal mechanisms of transcription control including RNA polymerase initiation and elongation and mRNA accumulation, transport and decay. Our approach extends to any discrete dynamic process with rare events and realistically limited data.