compute_alternative_SC: Compute super-cells using alternative algorithms behind....

In mariiabilous/SuperCellBM:

Description

Compute super-cells using alternative algorithms behind. Compute super-cells trying different number of neighbors for KNN, SNN implements in Seurat and also different clustering algorithms (i.e., current walktrap and Seurat-based Louvain)

Usage

compute_alternative_SC(
  counts,
  genes.use,
  meta.data = NULL,
  ge = NULL,
  N.comp = 10,
  k.seq = sort(c(5, 10, 50, 100, round(0.01 * ncol(counts)), round(0.05 *
    ncol(counts)))),
  res.seq = c(1, 2, 5, 10, 20, 30, 40, 50, 60, 70, 80),
  gamma.seq = NULL,
  return.singlecell.NW = FALSE,
  verbose = FALSE,
  common.gammas = TRUE,
  do.directed = c(T, F),
  group.singletons = FALSE,
  max.gamma = 150,
  ...
)

Arguments

counts	count matrix (genes x cells)
genes.use	a set of genes used in the original super-cell construction (output of SCimplify, field 'genes.use')
meta.data	meta data as an input to CreateSeuratObject
ge	log-normalized gene expression matrix (if computation is different from whar Seurat outputs)
k.seq	set of k to compute knn network
res.seq	set of resolutions to obtain different graining levels using louvain clustering in FindClusters
gamma.seq	graining levels to compute for walktrap clustering (is NULL, graining levels are retreived from the graining levels obtained in louvain clustering with res.seq resolutions)
return.singlecell.NW	whether to return single-cell network
common.gammas	whether to compute walktrap super-cells for both provided graining leveles (if providede) and obtained graning levels with louvain

Value

list of SuperCell -like structures with the first result layer corresponding to a NW type and the second layer corresponding to the clustering type (walktrap and louvain for the moment), the third layer is a graining level

mariiabilous/SuperCellBM documentation built on Jan. 28, 2022, 7:45 p.m.