README.md

In zky0708/2DImpute: 2DImpute: Imputing scRNA-seq data from correlations in both dimensions

2DImpute

An imputation method that recovers false zeros (known as dropouts) in single cell RNA sequencing (scRNA-seq) data. It features preventing excessive correction by predicting false zeros and imputing their values accurately by making use of the interrelationships between both genes and cells.

The users can refer to our paper published at Bioinformatics for a detailed description of the modeling and applications.

Installation

You can install 2DImpute from GitHub with:

# install.packages("devtools")
require(devtools)
install_github("weiyi-bitw/cafr")   # The 'cafr' R package is required for running 2DImpute.
install_github("zky0708/2DImpute")

Quick start

2DImpute takes as input a normalized scRNA-seq data matrix (genes in rows and cells in columns) with or without log-transformation. It outputs the imputed data in the same dimension and the same log-transformation. 1. Load the R package ("R2DImpute")

require(R2DImpute)

1. The imputation task can be done with one single function run_2DImpute:

res <- run_2DImpute(
  exprs,                      # The input expression matrix.
  t = 0.2,                    # Threshold set for dropout identification.
  genes = NULL,               # A vector of genes of which the zeros will get imputed. If it is NULL (default), all genes will be considered.
  k = 10,                     # Number of nearest neighbors.
  ncores = 1,                 # Number of cores to be used.
  log_input = TRUE            # Whether the input data is log-transformed. 
  return_J = FALSE,           # Whether to return the calculated pairwise Jaccard matrix between cells.
  return_attractors = FALSE,  # Whether to return the identified co-expressed gene attractor signatures.
  verbose = TRUE              # Whether to show the progress of imputation.
)
imputed_exprs <- res$imputed

For example:

# An example 10x scRNA-seq dataset containing 3,775 genes and 338 cells
data(ge_10x_sample)          

# Fast demonstration (only imputes zeros in genes 'XIST' and 'CD3D')
res <- run_2DImpute(
  exprs = ge_10x_sample, 
  genes = c('XIST', 'CD3D'), 
  ncores = 2
  )
imputed_exprs <- res$imputed

# Return identified co-expressed attractor signatures
res <- run_2DImpute(
  exprs = ge_10x_sample, 
  genes = c('XIST', 'CD3D'), 
  ncores = 2, 
  return_attractors = TRUE
  )
imputed_exprs <- res$imputed
attractors <- res$attractors

# Full imputation
res <- run_2DImpute(
  exprs = ge_10x_sample, 
  ncores = 2
  )
imputed_exprs <- res$imputed

zky0708/2DImpute documentation built on March 12, 2020, 3:13 a.m.