BEARscc-package: BEARscc (Bayesian ERCC Assesstment of Robustness of Single...

Description Details Author(s) References

Description

BEARscc is a noise estimation and injection tool that is designed to assess putative single-cell RNA-seq clusters in the context of experimental noise estimated by ERCC spike-in controls.

Details

Single-cell transcriptome sequencing data are subject to substantial technical variation and batch effects that can confound the classification of cellular sub-types. Unfortunately, current clustering algorithms don't account for this uncertainty. To address this shortcoming, we have developed a noise perturbation algorithm called BEARscc that is designed to determine the extent to which classifications by existing clustering algorithms are robust to observed technical variation.

BEARscc makes use of ERCC spike-in measurements to model technical variance as a function of gene expression and technical dropout effects on lowly expressed genes. In our benchmarks, we found that BEARscc accurately models read count fluctuations and drop-out effects across transcripts with diverse expression levels. Applying our approach to publicly available single-cell transcriptome data of mouse brain and intestine, we have demonstrated that BEARscc identified cells that cluster consistently, irrespective of technical variation. For more details, see the manuscript that is now available on bioRxiv.

Author(s)

David T. Severson <david_severson@hms.harvard.edu>

Maintainer: Benjamin Schuster-Boeckler <benjamin.schuster-boeckler@ludwig.ox.ac.uk>

References

Source code and README: <https://bitbucket.org/bsblabludwig/bearscc/overview> Associated preprint: <https://www.biorxiv.org/content/early/2017/06/05/118919>


seversond12/BEARscc documentation built on June 8, 2020, 8:04 a.m.