Skip to content

supporting code for the multinomial single cell RNA-Seq paper

License

Notifications You must be signed in to change notification settings

willtownes/scrna2019

Folders and files

NameName
Last commit message
Last commit date

Latest commit

aad379a · Jan 6, 2021

History

23 Commits
Oct 7, 2019
Nov 27, 2019
Oct 7, 2019
Feb 18, 2020
Jan 6, 2021
Mar 9, 2019
Mar 9, 2019
Oct 7, 2019
Jan 6, 2021
Mar 9, 2019

Repository files navigation

Feature Selection and Dimension Reduction for Single Cell RNA-Seq based on a Multinomial Model

DOI

This repository contains supporting code to facilitate reproducible analysis. For details see the Genome Biology publication. If you find bugs please create a github issue.

Please do not use this code for your own analyses! It is not updated. Better implementations are available in the following two R packages.

GLM-PCA (dimension reduction for generalized linear model likelihoods) is now available as a standalone R package. This method is highlighted in the paper as being suitable for single cell RNA-Seq data.

The scry R package contains functions for feature selection using deviance, computation of null residuals, and interfaces to apply these methods and GLM-PCA to Bioconductor objects such as SingleCellExperiment and SummarizedExperiment.

Authors

Will Townes, Stephanie Hicks, Martin Aryee, and Rafa Irizarry

Description of Repository Contents

algs

Implementations of dimension reduction algorithms

  • existing.R - wrapper functions for PCA, tSNE, ZINB-WAVE, etc
  • glmpca.R - placeholder file that just loads the glmpca package.

real

Analysis of various real scRNA-Seq datasets. The Rmarkdown files can be used to produce figures in the manuscript

real_benchmarking

Systematic assessment of clustering performance of a variety of normalization, feature selection, and dimension reduction algorithms using ground-truth datasets.

Downloadable table of results from assessments

util

Utility functions. Please consider using the updated versions of these functions via the scry R package.

  • clustering.R - wrappers for seurat clustering, model based clustering, and k-means
  • functions.R - Poisson and Binomial deviance and residuals functions, a function for loading 10x read counts from molecule information files.
  • functions_genefilter.R - convenience functions for gene filtering (feature selection) based on highly variable genes, highly expressed genes, and deviance.