Skip to contents

Ligand Receptor Analysis

Ligand_Receptor_Analysis.Rmd

To see what ligand-receptor channels are driving niche signals in our data, we developed niche-LR. Niche-LR uses the niche-net ligand-target potential matrix to find the top KK downstream targets for each ligand. The niche-DE T-statistics of these downstream genes are used to calculate ligand activity scores. The top MM ligands by ligand activity score and their corresponding receptors are then tested to confirm expression in the tissue. Niche-LR can be performed on spot resolution data using the function ‘niche_LR_spot’ and on single cell resolution data using the function ‘niche_LR_cell’. ‘niche_LR_spot’ takes in 9 arguments while ‘niche_LR_cell’ has one additional argument.

Arguments
  • object: Niche-DE object
  • ligand cell: The cell type that expresses the ligand
  • receptor cell: The cell type that expresses the receptor
  • ligand_target_matrix: A matrix that measures the association between ligands and their downstream target genes. The dimension should be #target genes by #ligands
  • lr-mat: A matrix that matches ligands with their corresponding receptors. This matrix should have two columns. The first will be ligands and the second will be the corresponding receptors
  • K: The number of downstream target genes to use when calculating the ligand activity score. Default value is 25.
  • M: The maximum number of ligands that can pass initial filtering. Default value is 50.
  • alpha: The level at which to perform the Benjamini Hochberg correction. Default value is 0.05.
  • truncation value: The value at which to truncate T statistics.
For ‘niche_LR_cell’, their is an additional parameter ‘alpha_2’ which refers to the null quantile to compare observed epression to. That is that we compare expression of the ligand and receptor to see if it is greater than the ‘alpha_2’ percentile of gene expression over all genes expressed in the ligand expressing or receptor expressing cell.The Default value is 0.5 (50th percentile).

We now perform ligand-receptor analysis to infer ligand-receptor interactions between tumor cells (ligand expressing cell) and fibroblasts (receptor expressing cell) on spot level data. The output will be a list of ligands and their corresponding receptors. Note that the ligand-target potential matrix refers to human genes.

data("niche_net_ligand_target_matrix")
data("ramilowski_ligand_receptor_list")
fibro_tumor_LR = niche_LR_spot(NDE_obj,ligand_cell = 'tumor_epithelial',receptor_cell = 'stromal',
ligand_target_matrix = niche_net_ligand_target_matrix,
lr_mat = ramilowski_ligand_receptor_list,K = 25,M = 50,alpha = 0.05,truncation_value = 3)
#preview output
head(fibro_tumor_LR)

The output should resemble a 3 column table of ligands,their corresponding receptors, and a list of the top 5 downstream niche-DE genes that drive the ligand potential scoring.