Using the Gene Ontology data objects

Daniel Greene

2021-02-10

ontologySimliarity comes with data objects encapsulating the GO (Gene Ontology) annotation of genes [1]:

• gene_GO_terms, a list of character vectors of term IDs of GO terms annotating each gene, named by gene,
• GO_IC, a numeric vector containing the information content of Gene Ontology terms based on frequencies of annotation in gene_GO_terms.

These data objects can be loaded in an R session using data(gene_GO_terms) and data(GO_IC) respectively. To process these objects, one can load the ontologyIndex package and a data object encapsulating the Gene Ontology.

library(ontologyIndex)
data(go)

library(ontologySimilarity)
data(gene_GO_terms)
data(GO_IC)

Users can simply subset the gene_GO_terms object to obtain GO annotation for their genes of interest, using a character vector of gene names. In this example, we'll use the BEACH domain containing gene family [2].

beach <- gene_GO_terms[c("LRBA", "LYST", "NBEA", "NBEAL1", "NBEAL2", "NSMAF", "WDFY3", "WDFY4", "WDR81")]

To see the names of the terms annotating a particular gene, the go ontology_index object can be used, using the term IDs to subset the name slot. For example, for "LRBA":

go$name[beach$LRBA]

##                       GO:0003674                       GO:0005764 
##             "molecular_function"                       "lysosome" 
##                       GO:0005783                       GO:0005794 
##          "endoplasmic reticulum"                "Golgi apparatus" 
##                       GO:0005886                       GO:0008150 
##                "plasma membrane"             "biological_process" 
##                       GO:0016021 
## "integral component of membrane"

The gene_GO_terms object contains annotation relating to all branches of the Gene Ontology, i.e. "cellular_component", "biological_process" and "molecular_function". If you are only interested in one branch - for example "cellular_component", you can use the ontologyIndex package's function intersection_with_descendants to subset the annotation.

cc <- go$id[go$name == "cellular_component"]
beach_cc <- lapply(beach, function(x) intersection_with_descendants(go, roots=cc, x)) 
data.frame(check.names=FALSE, `#terms`=sapply(beach, length), `#CC terms`=sapply(beach_cc, length))

##        #terms #CC terms
## LRBA        7         5
## LYST       14         2
## NBEA        6         4
## NBEAL1      3         2
## NBEAL2      4         2
## NSMAF      13         4
## WDFY3      13         8
## WDFY4       4         3
## WDR81      18         7

A pairwise gene semantic similarity matrix can be computed simply using the function get_sim_grid, and passing an ontology_index object, information content and annotation list as parameters (see ?get_sim_grid for more details). Here we plot the resulting similarity matrix using the paintmap package.

sim_matrix <- get_sim_grid(
    ontology=go, 
    information_content=GO_IC,
    term_sets=beach)

library(paintmap)
paintmap(colour_matrix(sim_matrix))

One can test whether a subset of genes is significantly similar as a group in the context of a larger collection by using the function get_sim_p_from_ontology to compute a p-value of similarity. For example here, we will compare the significance of the mean pairwise gene similarity within the BEACH group against randomly selected subsets of genes of the same size chosen from the gene_GO_anno set.

get_sim_p_from_ontology(
    ontology=go,
    information_content=GO_IC,
    term_sets=gene_GO_terms,
    group=names(beach)
)

## [1] 0.03484965