NBumi_ConvrtData: Convert Data to be suitable for NBumi
In M3Drop: Michaelis-Menten Modelling of Dropouts in single-cell RNASeq

Description Usage Arguments Details Value Examples

Recognizes a variety of R objects/classes and extracts expression matrices from them then converts that to a count matrix for input into NBumi functions.

1	NBumiConvertData(input, is.log=FALSE, is.counts=FALSE, pseudocount=1)

`input`	a matrix, data.frame or object
`is.log`	has the data been log-transformed? (assumes log-base 2 with pseudocount of 1)
`is.counts`	is the data raw unnormalized counts? (raw counts will be CPM normalized)
`pseudocount`	pseudocount added before log-transformation

You must have loaded the respective packages (in parentheses) into your namespace before running this function on the respective objects. Note that to maintain scalability sparse matrices will remain as such.

Supported classes/objects:

SCESet (scater <= 1.4.0): uses "counts" or if unavailable then "exprs"
SingleCellExperiment (scater >= 1.6.0): uses "counts", if unavailable then "logcounts", which is assumed to be log-normalized.
CellDataSet (monocle): uses "exprs", specify log/counts using arguments
ExpressionSet (Biobase): uses "exprs", specify log/counts using arguments
seurat (Seurat): uses "raw.data" as counts.

Matrix/Dataframe classes :

dgCMatrix (Matrix): specify log/counts using arguments
data.table (data.table): specify log/counts using arguments
DataFrame (S4Vectors): specify log/counts using arguments
AnnotatedDataFrame (Biobase): specify log/counts using arguments
matrix (base-r): specify log/counts using arguments
data.frame (base-r): specify log/counts using arguments

Counts are rounded up to integers if necessary, if counts are unavailable then this will attempt to convert log2 normalized expression to counts by de-logging, subtracting the pseudocount, and then un-normalizing by rescaling cells based on their relative number of detected genes, finally expression is rounded up to integers for use as counts.

A count matrix appropriate for input into NBumi functions.

	# Simulated raw count matrix:
	set.seed(42)
	counts <- matrix(rpois(200, lambda=3), ncol=10)
	input_counts <- NBumiConvertData(counts, is.counts=TRUE)

	# log normalized data frame
	lognorm <-log2( t(t(counts)/colSums(counts)*100)+1 )
	lognorm <- as.data.frame(lognorm)
	input_counts <- NBumiConvertData(lognorm)

	# Sparse matrix
	require("Matrix")
	counts <- Matrix(counts, sparse=TRUE)
	input_counts <- NBumiConvertData(counts, is.counts=TRUE)

	# SingleCellExperiment Object
	require("SingleCellExperiment")
	SCE <- SingleCellExperiment(assays=list(counts=counts))
	input_counts <- NBumiConvertData(SCE)

	# monocle Object
	require("monocle")
	obj <- suppressWarnings(newCellDataSet(as.matrix(lognorm)))
	input_counts <- NBumiConvertData(obj, is.log=TRUE)