adjacency_score: Ranks pairs of features by how localized they are in graph

In CamaraLab/AdjacencyScore: Computes a bivariate metric to rank pairs of features by how colocalized they are in a graph

Description

Given the adjacency matrix of a graph and a set of features on that graph, ranks given pairs of those features (f and g) by the equation f((e^cA-I)/c)g, which measures how much those features are colocalized in the graph. Calculates the p-value of this score by permuting the columns of the feature matrix.

Usage

adjacency_score(adj_matrix, f, f_pairs, c, num_perms = 1000, num_cores = 1,
  perm_estimate = F, groupings = F, verbose = T)

Arguments

adj_matrix	a (preferrably sparse) binary matrix of adjacency between the columns of f
f	a numeric matrix specifying one or more features defined for each node of the graph. Each column is a node of the graph and each row is a feature over the nodes.
f_pairs	a 2 column matrix where each row specifies the indices or names of a pair of features on which the score will be computed
c	constant used to determine width of diffusion, must be 0 <= c
num_perms	number of permutations used to build the null distribution for each feature. By default is set to 1000.
num_cores	integer specifying the number of cores to be used in the computation. By default only one core is used.
perm_estimate	boolean indicating whether normal distribution parameters should be determined from num_perms permutations to estimate the p-value. By default is set to FALSE.
groupings	boolean indicating whether features are binary and mutually exclusive indicated each point's inclusion in some group. Allows for p-value computation from a parameterized hypergeometric null distribution. By default is set to FALSE.
verbose	print time taken to create permutation matrices and compute adjacency score

CamaraLab/AdjacencyScore documentation built on May 23, 2021, 12:57 p.m.