runCONCLUS: Run CONCLUS in one click

In PolinaPavlovich/CONCLUS: ScRNA-seq workflow CONCLUS - from CONsensus CLUSters to a meaningful CONCLUSion.

Description

This function performs core CONCLUS workflow. It generates PCA and t-SNE coordinates, runs DBSCAN, calculates similarity matrices of cells and clusters, assigns cells to clusters, searches for positive markers for each cluster. The function saves plots and tables into dataDirectory.

Usage

runCONCLUS(sceObject, dataDirectory, experimentName,
  colorPalette = "default", statePalette = "default",
  clusteringMethod = "ward.D2", epsilon = c(1.3, 1.4, 1.5),
  minPoints = c(3, 4), k = 0, PCs = c(4, 6, 8, 10, 20, 40, 50),
  perplexities = c(30, 40), randomSeed = 42, deepSplit = 4,
  preClustered = F, orderClusters = FALSE, cores = 14,
  plotPDFcellSim = TRUE, deleteOutliers = TRUE,
  tSNEalreadyGenerated = FALSE, tSNEresExp = "")

Arguments

sceObject	a SingleCellExperiment object with your data.
dataDirectory	CONCLUS will create this directory if it doesn't exist and store there all output files.
experimentName	most of output file names of CONCLUS are hardcoded. experimentName will stay at the beginning of each output file name to distinguish different runs easily.
colorPalette	a vector of colors for clusters.
statePalette	a vector of colors for states.
clusteringMethod	a clustering methods passed to hclust() function.
epsilon	a parameter of fpc::dbscan() function.
minPoints	a parameter of fpc::dbscan() function.
k	preferred number of clusters. Alternative to deepSplit. A parameter of cutree() function.
PCs	a vector of first principal components. For example, to take ranges 1:5 and 1:10 write c(5, 10).
perplexities	a vector of perplexity for t-SNE.
randomSeed	random seed for reproducibility.
deepSplit	intuitive level of clustering depth. Options are 1, 2, 3, 4.
preClustered	if TRUE, it will not change the column clusters after the run. However, it will anyway run DBSCAN to calculate similarity matrices.
orderClusters	can be either FALSE (default) of "name". If "name", clusters in the similarity matrix of cells will be ordered by name.
cores	maximum number of jobs that CONCLUS can run in parallel.
plotPDFcellSim	if FALSE, the similarity matrix of cells will be saved in png format. FALSE is recommended for count matrices with more than 2500 cells due to large pdf file size.
deleteOutliers	whether cells which were often defined as outliers by dbscan must be deleted. It will require recalculating of the similarity matrix of cells. Default is FALSE. Usually those cells form a separate "outlier" cluster and can be easier distinguished and deleted later if necessary.
tSNEalreadyGenerated	if you already ran CONCLUS ones and have t-SNE coordinated saved You can set TRUE to run the function faster since it will skip the generation of t-SNE coordinates and use the stored ones. Option TRUE requires t-SNE coordinates to be located in your 'dataDirectory/tsnes' directory.
tSNEresExp	experimentName of t-SNE coordinates which you want to use. This argument allows copying and pasting t-SNE coordinates between different CONCLUS runs without renaming the files.

Value

A SingleCellExperiment object.

PolinaPavlovich/CONCLUS documentation built on May 10, 2019, 2:42 p.m.