Related Experiment Video
Updated: Jul 19, 2025
06:03
AMEBaS: Automatic Midline Extraction and Background Subtraction of Ratiometric Fluorescence Time-Lapses of Polarized Single Cells
Published on: June 23, 2023
498
Unsupervised removal of systematic background noise from droplet-based single-cell experiments using CellBender
Stephen J Fleming1,2, Mark D Chaffin3,4, Alessandro Arduini3,5
1Data Sciences Platform, Broad Institute of MIT and Harvard, Cambridge, MA, USA. sfleming@broadinstitute.org.
Nature Methods
|August 7, 2023
Summary
This study introduces CellBender, a deep learning tool to reduce background noise in droplet-based single-cell sequencing data. CellBender accurately removes noise, improving gene expression analysis and identifying uncaptured cell types.
Area of Science:
- Genomics
- Computational Biology
- Molecular Biology
Background:
- Droplet-based single-cell assays like scRNA-seq, snRNA-seq, and CITE-seq generate significant background noise.
- This noise, characterized by nonzero counts in cell-free droplets and off-target gene expression, can introduce batch effects and lead to inaccurate differential gene expression results.
Purpose of the Study:
- To develop a deep generative model for accurate noise quantification and removal in droplet-based single-cell assays.
- To implement this model into a scalable, open-source software package for broad accessibility.
Main Methods:
- Development of a deep generative model simulating noise generation in droplet-based assays.
- Implementation of the model into the CellBender software package.
- Validation using simulated and real-world single-cell sequencing datasets.
Main Results:
- The CellBender model accurately distinguishes cell-containing from cell-free droplets.
- It effectively learns and corrects for background noise, providing noise-free quantification.
- CellBender operates near the theoretical optimal denoising limit, as shown by simulations.
- Real-world data analysis demonstrated enhanced concordance with known gene expression patterns.
Conclusions:
- CellBender provides a robust solution for mitigating noise in droplet-based single-cell data.
- The software improves the reliability of gene expression analysis and facilitates the identification of potential data quality issues, such as degraded or uncaptured cell types.
