cluster_fcs_files: Cluster FCS files

In ParkerICI/scfeatures: Feature generation from single-cell data

Description

Cluster individual FCS files, using multiple CPU cores if possible

Usage

cluster_fcs_files(
  files.list,
  num.cores,
  col.names,
  num.clusters,
  asinh.cofactor,
  num.samples = 50,
  output.dir = ".",
  negative.values = "truncate",
  quantile.prob = 0.05
)

Arguments

files.list	The files to cluster
num.cores	Number of CPU cores to use
col.names	A vector of column names indicating which columns should be used for clustering
num.clusters	The desired number of clusters
asinh.cofactor	Cofactor for asinh transformation. If this is NULL no transformation is performed
num.samples	Number of samples to be used for the CLARA algorithm (see cluster::clara)
output.dir	The name of the output directory, it will be created if it does not exist
negative.values	How to deal with negative values in the data. If this is NULL negative values are left as is. Otherwise two options are possible: truncate: Negative values will be truncated (i.e. replaced with 0) shift: The data will be shifted so that only quantile.prob of the values for each channel will be truncated to 0. This option is useful in cases where the range of data significantly extends in the negatives, for instance due to compensation.
quantile.prob	Only used if negative.value is set to shift. The quantile of measurements that are going to be truncated to 0. For instance if this is 0.05, the data will be shifted so that only 5 percent of the values are negative and will be truncated to 0

Details

This function can produce two types of output:

• "file": Two files will be written, with names derived by appending ".clustered.txt" and ".all_events.rds" to the file names in files.list. The ".clustered.txt" file is a tab-separated table of median values for each cluster. The ".all_events.rds" file is an RDS file (readable with base::readRDS) containing a data.frame with all the rows in the original input file and an additional column called cellType, indicating cluster membership.

• "directory": In addition to the ".clustered.txt" file described above, this mode will create a folder called "clusters_data". This folder will contain a sub-folder for each input file, containing separate RDS files with the data in each cluster

Value

Returns either NULL or a try-error object if some error occurred during the computation

ParkerICI/scfeatures documentation built on April 10, 2021, 9:31 a.m.