What question did this study set out to answer?

To develop a statistical normalization framework that separates technical artifacts, such as cellular sequencing depth, from true biological heterogeneity in single-cell RNA sequencing data.

December 1, 2019Genome biologyOpen Access

Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression

View Full Paper

Ask AI

Bookmark

Authors

CHChristoph HafemeisterCommunity College of Rhode Island RSRahul SatijaBroad Institute

Discussion

Member takes

Implication

Methodological study demonstrates that regularized negative binomial regression removes technical noise from single-cell RNA-seq data, indicating improved accuracy in downstream genomic analyses.

Key Points

To develop a statistical normalization framework that separates technical artifacts, such as cellular sequencing depth, from true biological heterogeneity in single-cell RNA sequencing data.
Modeled unique molecular identifier counts using a generalized linear model with cellular sequencing depth as a covariate.
Applied regularized negative binomial regression by sharing parameter estimates across genes with similar abundance levels to prevent overfitting.
Calculated Pearson residuals to normalize count data and stabilize variance without heuristic pseudocounts or log-transformation.
Successfully removed cell-to-cell technical variation while preserving underlying biological heterogeneity across single-cell datasets.
Improved performance across standard downstream workflows, including variable gene selection, dimensional reduction, and differential expression analysis.
Packaged the framework into the open-source R package sctransform with integration into the Seurat single-cell analysis toolkit.

Cite This Study

Hafemeister et al. (2019) studied this question.

synapsesocial.com/papers/69d6aa8ffca0359822aa7f38 https://doi.org/10.1186/s13059-019-1874-1

View Full Paper

Ask AI

Bookmark

Implication

Methodological study demonstrates that regularized negative binomial regression removes technical noise from single-cell RNA-seq data, indicating improved accuracy in downstream genomic analyses.

Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression

Discussion

Member takes

Implication

Key Points

Cite This Study

Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider