R/enrichment_functions.R

In abodein/netOmics: Multi-Omics (time-course) network-based integration and interpretation

#' Get interaction from ORA enrichment analysis
#'
#' Returns results of an ORA analysis as an interaction graph
#'
#' @param query a vector (or a list) of character with the ID to perform
#'      the ORA analysis
#' @param sources (optional) a character in
#'    (GO, KEGG, REAC, TF, MIRNA, CORUM, HP, HPA, WP)
#' @param organism (optional) a character (default = 'hsapiens')
#' @param signif.value (optional) a logical, default = ''
#'
#' @return
#' a graph object (or list of graph) containing the interaction between
#' the query and the target terms.
#'
#' @seealso \code{\link[gprofiler2]{gost}}  \code{\link[gprofiler2]{gconvert}}
#'
#' @examples
#' query <- c('IL15', 'CDHR5', 'TGFA', 'C4B')
#' get_interaction_from_ORA(query,
#'                          sources = 'GO')
#'
#' query <- list('All' = c('IL15', 'CDHR5', 'TGFA', 'C4B'),
#'               'c1' = c('IL15', 'CDHR5', 'TGFA'))
#' get_interaction_from_ORA(query,
#'                          sources = 'GO')
#'
#' @importFrom gprofiler2 gost gconvert
#' @importFrom dplyr pull select filter
#' @export
get_interaction_from_ORA <- function(query,
                                     sources = "GO",
                                     organism = "hsapiens",
                                     signif.value = TRUE) {
  # validate query (char)
  if (is(query, "list")) {
    query <- lapply(query,
                    function(x)
                      check_vector_char(x,
                                        var.name = "'query '"))
    if (is.null(names(query))) {
      names(query) <- seq_along(query)
    }
  } else {
    query <- check_vector_char(query)
  }
  
  # check organism
  organism = check_vector_char(
    X = organism,
    X.length = 1,
    default = "hsapiens",
    var.name = "'organism' "
  )
  
  # check source
  sources <- match.arg(
    arg = sources,
    choices = c("GO", "KEGG", "REAC", "TF", "MIRNA", "CORUM", "HP",
                "HPA", "WP"),
    several.ok = FALSE
  )
  sources <-
    check_vector_char(sources, default = "GO")  # default value
  
  # check signif
  signif.value <- return_true_false(signif.value, default = TRUE)
  
  if (is(query, "list")) {
    res.ora <- list()
    term_map <- list()
    res.graph <- list()
    for (i in names(query)) {
      res.ora[[i]] <- get_ORA(query = query[[i]],
                              sources = sources,
                              organism = organism)
      
      term_map_tmp <- gprofiler2::gconvert(query = query[[i]],
                                           organism = organism,
                                           target = sources)
      
      target_id <- (
        res.ora[[i]] %>%
          dplyr::filter(significant == signif.value) %>%
          dplyr::pull(term_id)
      )
      
      term_map[[i]] <- term_map_tmp %>%
        dplyr::filter(target %in% target_id) %>%
        dplyr::select(input, target) %>%
        unique %>%
        na.omit()
      
      res.graph[[i]] <-
        igraph::graph_from_data_frame(term_map[[i]],
                                      directed = FALSE)
      res.graph[[i]] <- set_vertex_attr(
        graph = res.graph[[i]],
        name = "mode",
        index = term_map[[i]]$input,
        value = "core"
      )
      res.graph[[i]] <- set_vertex_attr(
        graph = res.graph[[i]],
        name = "mode",
        index = term_map[[i]]$target,
        value = "extended"
      )
      class(res.graph) <-
        c("list.interaction.igraph", "list.igraph")
    }
  } else {
    # query is not a list
    res.ora <- get_ORA(query = query,
                       sources = sources,
                       organism = organism)
    
    term_map_tmp <- gprofiler2::gconvert(query = query,
                                         organism = organism,
                                         target = sources)
    
    target_id <- (res.ora %>%
                    dplyr::filter(significant == signif.value) %>%
                    dplyr::pull(term_id))
    
    if (is.null(term_map_tmp)) {
      return(NULL)
    }
    term_map <- term_map_tmp %>%
      dplyr::filter(target %in% target_id) %>%
      dplyr::select(input, target) %>%
      unique %>%
      na.omit()
    
    res.graph <-
      igraph::graph_from_data_frame(term_map, directed = FALSE)
    res.graph <- set_vertex_attr(
      graph = res.graph,
      name = "mode",
      index = term_map$input,
      value = "core"
    )
    res.graph <- set_vertex_attr(
      graph = res.graph,
      name = "mode",
      index = term_map$target,
      value = "extended"
    )
    
    class(res.graph) <- c("interaction.igraph", "igraph")
    
  }
  return(res.graph)
}


#' ORA enrichment analysis
#'
#' Returns results of an ORA analysis
#'
#' @param query a vector of character, a lit of ID
#' @param sources a character or list of character
#' @param organism a character (default = 'hsapiens')
#'
#' @return
#' a data.frame containing the enrichment result
#'
#' @seealso \code{\link[gprofiler2]{gost}}
#'
#' @importFrom gprofiler2 gost
get_ORA <- function(query,
                    sources = NULL,
                    organism = "hsapiens") {
  if (is(query, "list")) {
    res <- list()
    for (i in names(query)) {
      ORA <- gprofiler2::gost(
        query = query[[i]],
        organism = organism,
        significant = FALSE,
        sources = sources,
        multi_query = FALSE
      )
      ORA.res <- ORA$result
      if (!is.null(ORA.res)) {
        ORA.res <- ORA.res %>%
          mutate(cluster = i) %>%
          dplyr::select(
            "cluster",
            "term_id",
            "source",
            "term_name",
            "p_value",
            "significant",
            "term_size",
            "query_size",
            "intersection_size",
            "precision",
            "recall"
          )
        res[[i]] <- ORA.res
      }
    }
    RES <- purrr::map_dfr(res, ~ .x)
    
  } else {
    ORA <- gprofiler2::gost(
      query = query,
      organism = organism,
      significant = FALSE,
      sources = sources,
      multi_query = FALSE
    )
    ORA.res <- ORA$result
    if (!is.null(ORA.res)) {
      ORA.res <- ORA.res %>%
        mutate(cluster = "All") %>%
        dplyr::select(
          "cluster",
          "term_id",
          "source",
          "term_name",
          "p_value",
          "significant",
          "term_size",
          "query_size",
          "intersection_size",
          "precision",
          "recall"
        )
      RES <- ORA.res
    } else {
      RES <- NULL
    }
  }
  return(RES)
}

#' Get GO info
#'
#' From a GO terms (GOID), return definition, ontology and term values
#'  from GO.db
#'
#' @param go a character, GO term
#'
#' @return
#' a data.frame with the following columns: 'GOID', 'DEFINITION',
#' 'ONTOLOGY', 'TERM'
#'
#' @import GO.db
#' @importFrom AnnotationDbi keytypes
get_go_info <- function(go) {
  res <- AnnotationDbi::select(
    x = GO.db,
    keys = go,
    keytype = "GOID",
    columns = keytypes(GO.db)
  )
  return(res)
}