rfCluster_row: rfCluster_row

In stela2502/BioData: The package allows to annotate a numeric data.table with col and row data

Description

This fucntion uses the RFclust.SGE to create fandomForest based unsupervised clusters on a subset of the data.

Default is on 200 cells using all (provided) genes with 500 forests and 500 trees per forest for 5 repetitions.

You are asked to give a k numer of expected clusters (better too many than too little), classifies the total

data using the 5 different unsupervised runs and all cluster ids from these runs are merged into the final cluster id.

This <summaryCol> will be part of the return objects samples table, together with a <usefulCol> where

all clusters with less than 10 cells have been merged into the 'gr. 0'.

The final results will be reported as new columns in the samples table containing the 'name'

Arguments

x	the single cells ngs object
email	your email to use together with the SGE option
SGE	whether to use the sun grid engine to calculate the rf grouping
rep	how many repetitions for the random forest grouping should be run (default = 5)
slice	how many processes should be started for each random forest clustering (default = 30)
bestColname	the column name to store the results in
k	the numer of expected clusters (metter more than to view)
subset	how many cells should be randomly selected for the unsupervised clustering (default = 200)
name	if you want to run multiple RFclusterings on e.g. using different input genes you need to specify a name (default ='RFclust')
nforest	the numer of forests to grow for each rep (defualt = 500)
ntree	the numer of trees per forest (default = 500)
settings	slurm settings list(A, t and p) which allow to run the rf clustering on a slurm backend
ids	the ids for a subset of genes to be analyzed (default NULL)

Value

a SingleCellsNGS object including the results and storing the RF object in the usedObj list (bestColname)

stela2502/BioData documentation built on Feb. 23, 2022, 5:47 a.m.