explore_clusters: explore_clusters

In charles-bernard/mage: Mine Associated Gene Expressions (from single-cell Rna Seq data)

Description

Given a table of scores and its partition into k clusters, the function explore_clusters proposes to explore the types of associations lying within each cluster.

To do so, the function measures the dispersion of the associations of each cluster in the space of the scores and depicts this heterogeneity as a tree structure (using ElPiGraph.R package). For instance, nodes along a same branch will be expected to exhibit similar types of associations.

Then, the function will visit the tree of each cluster to create an arborescence of directories starting from the root (the root is defined as the branch which is the closest to the centroid of the table of scores) and expanding from adjacent branches to adjacent branches.

Inside each directory corresponding to a branch (or a branching point), the function will produce the scatterplots of a few sampled assocations in order for the user to get a glimpse of what type of associations each branch is related to.

By visiting this arborescence of directories and visualizing some of the scatterplots of the associations each branch is related to, it enables one to somehow "wander" along a cluster's tree and appreciate the "evolution" of the "shape/type" of the associations from branches to branches

Usage

explore_clusters(scores_tab, partition, gene_exp_mat,
  sampling_method = "downsampling", target_ratio = 0.3, target_nb = NULL,
  output_directory = ".")

Arguments

scores_tab	data.table; the table of scores
partition	categorial vector; the partition of the scores_tab into k clusters. length(partition) must be equal to nrow(scores_tab)
gene_exp_mat	matrix: the gene expression matrix used for computing the scores (rownames corresponding to genes and colnames to cells)
sampling_method	method to be used for selecting a few associations in each branch. Methods available: "downsampling" "lowsampling" "upsampling" "centroids" "medoids" Method recommanded: "downsampling"
target_ratio	number; fraction of associations to be sampled in each tree's node . Argument target_ratio will be ignored if argument target_nb is provided
target_nb	number of associations to be sampled in each tree's node
output_directory	where to create the arborescence of directories for each cluster

charles-bernard/mage documentation built on May 14, 2019, 2 a.m.