BurstDECONV: a signal deconvolution method to uncover mechanisms of transcriptional bursting in live cells

Monitoring transcription in living cells gives access to the dynamics of this complex fundamental process. It reveals that transcription is discontinuous, whereby active periods (bursts) are separated by one or several types of inactive periods of distinct lifetimes. However, decoding temporal fluctuations arising from live imaging and inferring the distinct transcriptional steps eliciting them is a challenge. We present BurstDECONV, a novel statistical inference method that deconvolves signal traces into individual transcription initiation events. We use the distribution of waiting times between successive polymerase initiation events to identify mechanistic features of transcription such as the number of rate-limiting steps and their kinetics. Comparison of our method to alternative methods emphasizes its advantages in terms of precision and flexibility. Unique features such as the direct determination of the number of promoter states and the simultaneous analysis of several potential transcription models make BurstDECONV an ideal analytic framework for live cell transcription imaging experiments. Using simulated realistic data, we found that our method is robust with regards to noise or suboptimal experimental designs. To show its generality, we applied it to different biological contexts such as Drosophila embryos or human cells.

Automated summary

Monitoring transcription in living cells reveals that it is discontinuous with active periods separated by inactive periods of distinct lifetimes. However, decoding temporal fluctuations and understanding the underlying transcriptional steps is challenging. BurstDECONV is a statistical inference method that can identify individual transcription initiation events and extract mechanistic features of transcription. Compared to alternative methods, BurstDECONV has advantages in terms of precision and flexibility, making it an ideal framework for live cell transcription imaging experiments. It is robust to noise and applicable to different biological contexts.