calcDM: Calculate a diffusion map
In farrellja/URD: URD

Description Usage Arguments Details Value See Also Examples

URD uses the R package destiny to calculate a diffusion map.

calcDM(
  object,
  genes.use = object@var.genes,
  cells.use = NULL,
  knn = NULL,
  sigma.use = NULL,
  n_local = 5:7,
  distance = c("euclidean", "cosine", "rankcor"),
  density.norm = T,
  dcs.store = 200,
  verbose = T
)

`object`	An URD object
`genes.use`	(Character vector) Genes to use (default is variable genes, NULL uses all genes)
`cells.use`	(Character vector) Cells to use (default is NULL, which uses all cells)
`knn`	(Numeric) Number of nearest neighbors to use (NULL takes a guess)
`sigma.use`	(NULL, numeric, or "local") Kernel width to use for the diffusion map (NULL uses destiny's global auto-detection procedure, and "local" determines a sigma for each cell based on the distance to its nearest neighbors.)
`n_local`	(Numeric) If `sigma.use="local"`, the distance to the n-local(th) nearest neighbors are used for determining the kernel width for each cell.
`distance`	(Character) Distance metric to use for determining transition probabilities.
`density.norm`	(Logical) If TRUE, use density normalization
`dcs.store`	(Numeric) How many diffusion components to store
`verbose`	(Logical) Report determined values for auto-detected parameters?

The diffusion map is stored in the @dm slot of the URD object, and the transition probabilities (@dm@transitions) are used extensively downstream in URD's pipeline. Important parameters to vary while choosing a diffusion map to use are the number of nearest neighbors considered (the square root of the number of cells is a good starting point, or NULL will trigger destiny to try to determine a good starting point), and the sigma of the Gaussian used to transform cells' distance.

An URD object with the diffusion map stored in @dm.

DiffusionMap

# Allow destiny to determine most parameters on its own
object <- calcDM(object, knn=NULL, sigma.use=NULL)

# Refine the diffusion map by adjusting parameters manually
object <- calcDM(object, knn = 200, sigma.use = 8)