GetSignalingPartners: Compute Cell-cell interaction probability

In mkarikom/RSoptSC: Low-Rank Factorization for Subspace-clustering

Compute Cell-cell interaction probability

Description

We can have a situation where ligand/receptor expression is low, but target genes (repressors and activators) are highly expressed. In this case we will get a false positive for P_{i,j}. In order to correct for this, we introduce normalizing coefficients for the relationship between \alpha_{i,j} and \beta / \gamma. When \alpha is low, then K and D will decrease rapidly with increasing \beta and \gamma, penalizing the resulting increase in P (which is increasingly likely to be a false positive, given the disparity between \alpha and \beta,\gamma) K_{i,j} = \frac{\alpha_{i,j}}{\alpha_{i,j} + \beta_{i,j}} Note that activated genes are required under this model.

Usage

GetSignalingPartners(data, ids, pathway, normalize_aggregate = TRUE)

Arguments

data	a matrix of expression values for each cell (rows) and gene (columns)
ids	a vector of gene ids
pathway	a data frame with "ligands", "receptors", "direction", and "targets"
normalize_aggregate	whether or not to normalize the P_agg matrix by dividing each row by its sum default is true

Value

a list containing:

P	a list of the cell-cell signaling probabilities for all ligand/receptor pairs
P_agg	the aggregate matrix of all ligand/receptor pairs

mkarikom/RSoptSC documentation built on May 10, 2023, 1:10 a.m.