get_collectri is hard coded for human

[karJac]

Hi,

After installing the decoupleR package from Bioconductor on R version 4.3.2, I encountered an issue with the function decoupleR::get_collectri(). Although the function's description suggests that it supports both human and mouse datasets, its implementation appears to be hardcoded exclusively for human data.

We can see in the code that parameter "organism" is not being caleld inside OmnipathR::collectri(). The default parameter for "organism" in OmnipathR::collectri() is human, thus the function is hardcoded for "human".

Here is the code of get_collectri function with correct calling of "organism" parameter:

get_collectri <- function(organism = "human", split_complexes = FALSE){
  if (organism == "human"){
    organism <- "9606"
  } else if (organism == "mouse"){
    organism <- "10090"
  } else {
    stop("Interactions are available only for human and mouse!")
  }

  collectri <- OmnipathR::collectri(organism=organism) #default organism is human
  cols <- c("source_genesymbol", "target_genesymbol", "is_stimulation", 
            "is_inhibition")
  collectri_interactions <- collectri[!stringr::str_detect(collectri$source, 
                                                           "COMPLEX"), cols]
  collectri_complex <- collectri[stringr::str_detect(collectri$source, 
                                                     "COMPLEX"), cols]
  if (!split_complexes) {
    collectri_complex <- collectri_complex %>% dplyr::mutate(source_genesymbol = dplyr::case_when(stringr::str_detect(source_genesymbol, 
                                      "JUN") | stringr::str_detect(source_genesymbol, "FOS") ~ 
                    "AP1", stringr::str_detect(source_genesymbol, "REL") | 
                    stringr::str_detect(source_genesymbol, "NFKB") ~ 
                    "NFKB"))
  }
  collectri <- base::rbind(collectri_interactions, collectri_complex) %>% 
    dplyr::distinct(.data$source_genesymbol, .data$target_genesymbol, 
                    .keep_all = TRUE) %>% dplyr::mutate(weight = dplyr::case_when(is_stimulation == 
                         1 ~ 1, is_stimulation == 0 ~ -1)) %>% dplyr::select(.data$source_genesymbol, 
                                    .data$target_genesymbol, .data$weight) %>% dplyr::rename(source = .data$source_genesymbol, 
                                          target = .data$target_genesymbol, mor = .data$weight, 
                                    )
  return(collectri)
}

Now it seems to work properly.

Hope that my post will be helpful for you and thank you for creating this package! Karol

[deeenes]

Hi Karol,

This has been fixed long time ago, I recommend to update the package from this git repo:

library(remotes)
remotes::install_github('saezlab/decoupleR')

I'll check why the update is not included in the latest BioC release.

Best,

Denes

[karJac]

It seems that BiocManager::install("decoupleR") installed the version 2.6.0, instead of the latest 2.8.0, because BiocManager was outdated (3.16 instead of 3.18). My bad then, sorry for the confusion.