Calculate GeneWalk test statistics — rs_gene_walk

Pools all null distributions into one, then for each observed embedding rep: computes p-values, per-gene FDR, and global FDR. Finally aggregates across reps with geometric mean and 95\

Matches the paper: "we repeat the above-described network representation learning and significance testing procedures of the GWN nreps_graph times and provide the mean and 95\

Usage

rs_gene_walk_test(
  gene_embds_list,
  pathway_embds_list,
  null_similarities,
  connected_pathways,
  verbose
)

Arguments

gene_embds_list: List of n_graph gene embedding matrices (n_genes x dim).
pathway_embds_list: List of n_graph pathway embedding matrices (n_pathways x dim).
null_similarities: List of n_perm numeric vectors (null cosine similarities to be pooled).
connected_pathways: List. Per-gene integer vectors of connected pathway indices (1-indexed).
verbose: Controls verbosity.

Value

A list with per-pair statistics.