NBumiConvertData {M3Drop} | R Documentation |
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.
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 DataTable (S4Vectors) : 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)