ORA, GSEA and enrichment map

In this vignette we will show how to use perform Over-Representatin Analysis (ORA) and Gene Set Enrichment Analysis (GSEA) with NPATools.

ORA and GSEA

ORA is performed by means of the function ora(), which requires a list of one or more gene vectors (wb), the reference universe and a list of gene sets to test (gsl):

ora_res <- ora(wb = wb, universe = unverse, gsl = gsl)

GSEA is performed by means of the function gsea(), which requires a matrix with one or more columns of gene-related statistics (rl) and a list of gene sets to test (gsl):

gsea_res <- gsea(rl=rl, gsl = gsl)

By default, each column of rl is sorted in decreasing order.

Visualization

The functions plot_ora_heatmap() and plot_gsea_heatmap() draw heatmaps with results of ORA and GSEA:

plot_ora_heatmap(ora_res, cluster_columns = F, cluster_rows = F, a = .5, na_col = "gold")
plot_gsea_heatmap(gsea_res = gsea_res, cluster_columns = F, cluster_rows = F, a = .5, na_col = "gold")

Inter-operability

We provide the functions ora2enrich() and gsea2enrich() to obtain instances of “enrichResult” class of the package clusterProfiler:

ora_enrich <- ora2enrich(ora_res, gsl = gsl, wb = wb, universe = unverse)
gsea_enrich <- gsea2enrich(gsea_res, rl = rl, gsl = gsl)

Functional clustering and enrichment map

The following procedure calculates an enrichment map and the corresponding gene set clusters. These can be used, for example, to create visualizations where gene-associated values are clustered by gene membership in similar gene sets. This procedure is often called “functional clustering” because gene sets defines gene functions. The two required inputs are:

• gene_set_scores: a named vector of gene set scores;
• gene_set_list: a named list of gene set vectors.

The first step is the calculation of all-pairs similarities among gene sets:

gene_set_sim <- calc_gs_sim(gene_set_list)

The function is based on proxy::simil() and, by default gene_set_sim(), uses “Simpson” similarity measure, also known as overlap coefficient; a list of all available measures can be obtained using summary(proxy::pr_DB) and more information about a particular measure can be shown by proxy::pr_DB$get_entry("Simpson") (see ?proxy::simil() for further details). Once we have all-pairs similarities we can create the corresponding enrichment map (which requires gene set sizes):

gs_size <- setNames(lengths(gene_set_list), names(gene_set_list))
em_res <- enrichment_map(gs_scores=gene_set_scores, gene_set_sim=gene_set_sim, gs_size=gs_size)

A relevant argument is min_sim (by default 0.2), which defines the minimum threshold for similarity values that will be considered (the others are set to zero). A data frame of gene sets, showing their cluster, score and size can be retrieved using:

gene_set_clusters_df <- get_gene_set_clusters(em_res)

A named list of gene set vectors that defines gene-cluster membership can be obtained by:

clust_genes <- get_cluster_genes(gene_set_clusters=gene_set_clusters_df, gene_set_list=gene_set_list)