In aertslab/SCENIC: SCENIC (Single Cell rEgulatory Network Inference and Clustering)

# Suppress loading messages when building the HTML
suppressPackageStartupMessages({
  library(SCENIC)
  library(data.table)
  library(GSEABase)
  library(AUCell)
  library(SCopeLoomR)
})

This vignette is deprecated. pySCENIC (VSN) results are now saved into a loom file by default. See section Loading results from a .loom file from the main vignette.

pySCENIC (https://pyscenic.readthedocs.io/) provides a faster implementation of SCENIC (in Python) that can be easily paralelized with Dask (e.g. to take advantage distributed systems). The results of pySCENIC can easily be imported into R.

The main results of SCENIC analysis are stored in the loom file:

• Regulons

• Regulon activity (AUC matrix and thresholds)

• Embeddings (e.g. t-SNE and UMAP on the regulon activity)

The motif enrichment analysis and co-expression modules (e.g. GRNBoost/GENIE3 output) are stored in independent text files (mostly due to their bigger size).

# Output directories (adjust to yours):
pyScenicDir <- "pySCENIC_example/out"

library(SCENIC)

Reading the loom file

SCopeLoomR provides functions to import the regulons, AUC, and embeddings from the loom file:

library(SCopeLoomR)

pyScenicLoomFile <- file.path(pyScenicDir, "SCENIC.loom")
loom <- open_loom(pyScenicLoomFile, mode="r")

# Read information from loom file
# Regulons:
regulons_incidMat <- get_regulons(loom) # as incid matrix
regulons_motif <- regulonsToGeneLists(regulons_incidMat) # convert to list

regulons_ChIP <- regulonsToGeneLists(get_regulons(loom, attrName = "TrackRegulons"))

# Regulon AUC and thresholds
regulonsAUC <- get_regulonsAuc(loom)
regulonsAucThresholds <- get_regulonThresholds(loom)

# Embeddings (tsne/umap)
embeddings <- get_embeddings(loom)

In addition, the loom file also stores the expression matrix, and allows to save information about the cells (e.g. cell annotations), or results from previous clustering analyses:

exprMat <- get_dgem(loom)
# cellInfo <- get_cellAnnotation(loom) # will also contain AUC values, etc... you can filter them out
clusterings <- get_clusterings_withName(loom)

close_loom(loom)

Motif enrichment analysis

The motif enrichment results provided by pySCENIC are in a slightly different format than the table provided by RcisTarget (e.g. R implementation of SCENIC) but they are equivalent. This table can be read from the text file and visualized/explored in a similar way:

sampleName <- "yourSampleName"
# Read table
motifsDf <- data.table::fread(file.path(pyScenicDir, "out/scenic/",sampleName,"/cistarget/",sampleName,"__reg_mtf.csv"), header = T, sep="\t")
maxRows <- 20 # (low value only for the tutorial)

# Visualize
tableSubset <- motifsDf[TF=="Dlx5"]
tableSubset <- tableSubset[1:maxRows,] 
colsToShow <- colnames(motifsDf)[-c(2,9:11)]
viewMotifs(tableSubset, colsToShow=colsToShow)

GRNBoost/GENIE3 results

GRNBoost_linkList <- importArboreto(file.path(pyScenicDir,  "adjacencies.tsv"))
head(GRNBoost_linkList)

aertslab/SCENIC documentation built on April 7, 2024, 10 a.m.