repSample: Downsampling and resampling of immune repertoires

In immunomind/immunarch: Bioinformatics Analysis of T-Cell and B-Cell Immune Repertoires

Downsampling and resampling of immune repertoires

Description

Sample (downsample) repertoires using different approches.

Usage

repSample(
  .data,
  .method = c("downsample", "resample", "sample"),
  .n = NA,
  .prob = TRUE
)

Arguments

.data	The data to be processed. Can be data.frame, data.table, or a list of these objects. Every object must have columns in the immunarch compatible format. immunarch_data_format Competent users may provide advanced data representations: DBI database connections, Apache Spark DataFrame from copy_to or a list of these objects. They are supported with the same limitations as basic objects. Note: each connection must represent a separate repertoire.
.method	Character. Name of a sampling method. See "Details" for more details. Default value is "downsample" that downsamples the repertoires to the number of clones (i.e., reads / UMIs) that the smallest repertoire has, if user doesn't set any value to the ".n" argument.
.n	Integer. Number of clones / clonotypes / reads / UMIs to choose, depending on the method. Set NA to sample repertoires to the size of the smallest repertoire in the ".data".
.prob	Logical. If TRUE then samples the clonotypes with probability weights equal to their number of clones. Used only if ".method" is "sample".

Details

If .method is "downsample" then repSample chooses .n clones (not clonotypes!) from the input repertoires without any probabilistic simulation, but exactly computing each choosed clones. Such approach is is more consistent and biologically pleasant than an output from the function if .method is "resample".

If .method is "resample" then repSample uses multinomial distribution to compute the number of occurences for each cloneset. then it removes zero-number clonotypes and return the resulting data frame. Probabilities for rmultinom for each cloneset is a percentage of this cloneset in the "Proportion" column. It's a some sort of simulation of how clonotypes are chosen from the organisms.

if .method is "sample" then repSample chooses .n clonotypes (not clones!) randomly. Depending on the .prob argument, the function chooses clonotypes either according to their size (if .prob is TRUE, by default), or each clonotype has an equal chance to be choosed (if .prob is FALSE). Note that sampling is done without replacing.

Value

Subsampled immune repertoire or a list of subsampled immune repertoires.

See Also

rmultinom, clonal_proportion

Examples

data(immdata)
# Downsampling to 1000 clones (not clonotypes!)
tmp <- repSample(immdata$data[[1]], .n = 1000)
sum(tmp$Clones)

# Downsampling to 1000 clonotypes
tmp <- repSample(immdata$data[[1]], "sample", .n = 1000)
nrow(tmp)

# Downsampling to the smallest repertoire by clones (not clonotypes!)
tmp <- repSample(immdata$data[c(1, 2)])
sum(tmp[[1]]$Clones)
sum(tmp[[2]]$Clones)

# Downsampling to the smallest repertoire by clonotypes
tmp <- repSample(immdata$data[c(1, 2)], "sample")
nrow(tmp[[1]]$Clones)
nrow(tmp[[2]]$Clones)

immunomind/immunarch documentation built on March 20, 2024, 12:01 p.m.