create a nearest neighbour (NN) network
createNearestNetwork(
gobject,
type = c("sNN", "kNN"),
dim_reduction_to_use = "pca",
dim_reduction_name = "pca",
dimensions_to_use = 1:10,
genes_to_use = NULL,
expression_values = c("normalized", "scaled", "custom"),
name = "sNN.pca",
return_gobject = TRUE,
k = 30,
minimum_shared = 5,
top_shared = 3,
verbose = T,
...
)giotto object
sNN or kNN
dimension reduction method to use
name of dimension reduction set to use
number of dimensions to use as input
if dim_reduction_to_use = NULL, which genes to use
expression values to use
arbitrary name for NN network
boolean: return giotto object (default = TRUE)
number of k neighbors to use
minimum shared neighbors
keep at ...
be verbose
additional parameters for kNN and sNN functions from dbscan
giotto object with updated NN network
This function creates a k-nearest neighbour (kNN) or shared nearest neighbour (sNN) network based on the provided dimension reduction space. To run it directly on the gene expression matrix set dim_reduction_to_use = NULL.
See also kNN and sNN for more information about
how the networks are created.
Output for kNN:
from: cell_ID for source cell
to: cell_ID for target cell
distance: distance between cells
weight: weight = 1/(1 + distance)
Output for sNN:
from: cell_ID for source cell
to: cell_ID for target cell
distance: distance between cells
weight: 1/(1 + distance)
shared: number of shared neighbours
rank: ranking of pairwise cell neighbours
For sNN networks two additional parameters can be set:
minimum_shared: minimum number of shared neighbours needed
top_shared: keep this number of the top shared neighbours, irrespective of minimum_shared setting
data(mini_giotto_single_cell)
mini_giotto_single_cell <- createNearestNetwork(gobject = mini_giotto_single_cell,
dimensions_to_use = 1:3, k = 3)