Global transcription regulation revealed from dynamical correlations in time-resolved single-cell RNA-sequencing

Abstract

Single-cell transcriptomics reveals significant variations in the transcriptional activity across cells. Yet, it remains challenging to identify mechanisms of transcription dynamics from static snapshots. It is thus still unknown what drives global transcription dynamics in single cells. We present a stochastic model of gene expression in growing and dividing cells that harnesses temporal dimensions of single-cell RNA-sequencing through metabolic labelling protocols and cell cycle reporters. We develop a parallel and highly scalable Approximate Bayesian Computation method that corrects for technical variation and accurately quantifies absolute burst frequency, burst size and degradation rate at a transcriptome-wide scale. Using Bayesian model selection, we reveal scaling between transcription rates and cell size and unveil waves of gene regulation across the cell cycle-dependent transcriptome. Our study shows that stochastic modelling of dynamical correlations identifies global mechanisms of transcription regulation. We illustrate how mechanistic modelling and amortised inference can efficiently integrate distributions across time, modalities and conditions. Our framework thus provides a versatile tool to uncover how global transcriptional dynamics is orchestrated genome-wide in single cells.

Transcriptome-wide analysis of cell cycle-dependent bursty gene expression from single-cell RNA-seq data using mechanistic model-based inference

“…The error in the burst frequency was actually larger than if we used a model that did not at all account for extrinsic noise in the transcription rate. Hence, our results suggest that several recently developed approaches using models that account for extrinsic noise in the transcription rate but assume a fixed number of gene copies per cell (typically one) to infer the burst parameters from non-allele-specific single-cell sequencing data [44, 45, 56] should be used with great care. Instead, the estimation of ratios of the burst parameters in G1 and G2/M from cell-cycle phase specific scRNA-seq data, using mechanistic models that account for gene copy number variation across the cell cycle, is generally a far more robust inference strategy.…”

“…As previously mentioned, DeepCycle assigns a normalised cell age θ to each cell, which varies between 0 and 1. For a cell in generation i, θ can be understood as equal to t/T i , where T i is the cell-cycle duration in this generation and t is the true cell age (which varies between 0 and T i ); hence in this method there is no implicit assumption of a fixed cell-cycle duration as in other inference methods [56]. This is important because cell-cycle duration variability is a significant contributor to the magnitude of gene expression noise [58].…”

“…For instance, errors of this type are made when fitting a model with one gene copy to non-allele-specific data in eukaryotic cells, such as in Ref. [56], or when we previously fit Model II to the scRNA-seq mouse embryonic stem cell data used in this paper. In particular, the errors due to inference using Model II stem from the fact that it assumes 2 gene copies but many cells in the population are in the G2/M phase and instead have 4 gene copies.…”

“…This trifarious correction is missing from the literature. Current methods of inferring burst parameters from scRNA-seq data either ignore/cannot properly account for extrinsic and technical noise [42, 43] or correct only for extrinsic noise [19] or correct for certain types of extrinsic noise and technical noise but not for the variation in gene copy number during the cell-cycle [44, 45, 48, 56].…”