Related Experiment Video
Updated: Mar 18, 2026
10:36
Rare Event Detection Using Error-corrected DNA and RNA Sequencing
Published on: August 3, 2018
12.7K
From noise to models to numbers: Evaluating negative binomial models and parameter estimations in single-cell
Yiling Wang1, Zhanpeng Shu2, Zhixing Cao1,3
1State Key Laboratory of Bioreactor Engineering, East China University of Science and Technology, Shanghai, China.
Plos Computational Biology
|March 16, 2026
Summary
The Negative Binomial distribution approximates single-cell RNA sequencing (scRNA-seq) data well in specific gene expression parameter ranges, even without transcriptional bursting. Burst parameters are most informative when compared relatively.
Area of Science:
- Genomics
- Computational Biology
- Biostatistics
Background:
- The Negative Binomial (NB) distribution is commonly used for single-cell RNA sequencing (scRNA-seq) data, but its widespread applicability is not fully understood.
- Gene expression stochasticity is governed by kinetic parameters, influencing transcript count distributions.
Purpose of the Study:
- To investigate the relationship between gene expression kinetic parameters and the best-fit models (Beta-Poisson, NB, Poisson) for scRNA-seq data.
- To understand why the NB distribution is frequently the best fit for scRNA-seq data.
Main Methods:
- Employed a computationally efficient model selection technique.
- Simulated scRNA-seq data incorporating biological and technical noise.
- Mapped best-fit models to kinetic parameters of gene expression.
Main Results:
- The NB distribution approximates simulated data well within an intermediate range of normalized gene activation/inactivation rates.
- This NB-fitting range broadens with lower mean expression, higher technical noise, and larger sample sizes.
- Good NB fits do not exclusively indicate transcriptional bursting, and relative burst parameters are more reliable than absolute estimates.
Conclusions:
- The NB distribution's ubiquity in scRNA-seq data arises from its robustness across various gene expression parameter regimes.
- Biological noise significantly influences NB profiles in small sample sizes.
- While absolute burst parameters have high relative errors, relative ranking of burst frequency remains accurate, informing gene expression dynamics.
