data-raw/alias-table.R

In MSKCC-Epi-Bio/gnomeR: Wrangle and analyze IMPACT and TCGA mutation data

# # This script creates and saves the IMPACT Gene Info Data Frame within the package

library(cbioportalR)
library(tidyverse)
library(gnomeR)

set_cbioportal_db("public")

# Functions --------------------------------------------------------------------

# function clean gene names
clean_genes <- function(impact_plat, name) {

  impact_plat %>%
    mutate(gene = str_replace_all(gene, "-", "."),
              platform = name) %>%
    filter(gene != "Tiling") %>%
    distinct()
}

# a function to get aliases from cbioportal API
get_alias <- function(hugo_symbol) {
  url_path = paste0("genes/", hugo_symbol, "/aliases")
  res <- cbioportalR::cbp_api(url_path)

  res <- res$content
  unlist(res)
}


# IMPACT ------------------------------------------------------------------


# * Get IMPACT genes -----------------------------------------------------------

api_230 <- get_gene_panel(panel_id = "IMPACT230") %>%
  transmute(gene = hugoGeneSymbol,
            entrez_id = entrezGeneId)

api_279 <- get_gene_panel(panel_id = "IMPACT279") %>%
  transmute(gene = hugoGeneSymbol,
            entrez_id = entrezGeneId)

api_300 <- get_gene_panel(panel_id = "IMPACT300") %>%
  transmute(gene = hugoGeneSymbol,
            entrez_id = entrezGeneId)

api_341 <- get_gene_panel(panel_id = "IMPACT341") %>%
  transmute(gene = hugoGeneSymbol,
         entrez_id = entrezGeneId)

api_410 <- get_gene_panel(panel_id = "IMPACT410") %>%
  transmute(gene = hugoGeneSymbol,
         entrez_id = entrezGeneId)

api_468 <- get_gene_panel(panel_id = "IMPACT468") %>%
  transmute(gene = hugoGeneSymbol,
         entrez_id = entrezGeneId)

api_505 <- get_gene_panel(panel_id = "IMPACT505") %>%
  transmute(gene = hugoGeneSymbol,
            entrez_id = entrezGeneId)

# heme panels
api_heme_400 <- get_gene_panel(panel_id = "IMPACT-HEME-400") %>%
  transmute(gene = hugoGeneSymbol,
            entrez_id = entrezGeneId)

api_heme_468 <- get_gene_panel(panel_id = "IMPACT-HEME-468") %>%
  transmute(gene = hugoGeneSymbol,
            entrez_id = entrezGeneId)

l_api <- list(
          "api_230" = api_230,
          "api_279" = api_279,
          "api_300" = api_300,
          "api_341" = api_341,
          "api_410" = api_410,
          "api_468" = api_468,
          "api_505" = api_505,
          # heme panels
          "api_heme_400" = api_heme_400,
          "api_heme_468" = api_heme_468)

# create data frame of genes
# impact_genes <- map2_df(l_api, names(l_api),
#                         ~clean_genes(impact_plat = .x,
#                                        name = .y))

impact_genes <- l_api %>% bind_rows()

impact_genes <- impact_genes %>%
  select(gene, entrez_id) %>%
  unique()


# * Get Gene Entrez IDs  --------------------------------------------------------

# # API call to get a list of genes and entrez ids
all_genes <- get_genes()

all_genes <- all_genes %>%
  filter(type == "protein-coding") %>%
  janitor::clean_names() %>%
  rename("hugo_symbol" = hugo_gene_symbol)

# * Get Aliases  --------------------------------------------------------

# get all aliases for impact genes
impact_alias <- impact_genes %>%
  select(gene) %>%
  dplyr::mutate(alias = purrr::map(gene,
                                   ~tryCatch(get_alias(.x),
                                             error = function(e) NA_character_)))
# unnested long version
impact_alias_table <- impact_alias %>%
  unnest(cols = c(alias),
         keep_empty = TRUE)

impact_alias_table %>%
  filter(gene == alias)


impact_alias_table <- impact_alias_table %>%
  filter(gene != alias)

# This may not be needed anymore, but leave for now
impact_alias_table$gene[map_lgl(impact_alias_table$gene, ~.x %in%
                                  impact_alias_table$alias)] %>%
  unique()

impact_alias_table <- impact_alias_table %>%

  # I don't think these are aliases even though they came up. They usually only come up one way
  # I think they are separate genes although I am not sure
  filter(!(gene == "SOS1" & alias == "HGF")) %>%
  filter(!(gene == "HRAS" & alias == "KRAS")) %>%
  filter(!(gene == "TP53BP1" & alias == "TP53")) %>%
  filter(!(gene == "EZH2" & alias == "EZH1")) %>%

  # are MLL2 and MLL4 interchangeable?! - JJ email on 1/25/2021 says KMT2B = MLL4
  filter(!(gene == "KMT2D" & alias == "MLL4")) %>%
  filter(!(gene == "KMT2B" & alias == "MLL2")) %>%
  filter(!(gene == "ATRX" & alias == "RAD54L")) %>%
  distinct()

# check we've recoded all
impact_alias_table$gene[map_lgl(impact_alias_table$gene, ~.x %in% impact_alias_table$alias)] %>%
  unique()

impact_alias_table <- impact_alias_table %>%
  rename("hugo_symbol" = gene)

impact_alias_table <- impact_alias_table %>%
  left_join(., select(impact_genes, gene, entrez_id),
               by = c("hugo_symbol" = "gene")) %>%
  left_join(., select(all_genes, hugo_symbol, entrez_gene_id),
                       by = c("alias" = "hugo_symbol")) %>%
             rename(alias_entrez_id = entrez_gene_id)

impact_alias_table <- impact_alias_table %>%
  distinct()


usethis::use_data(impact_alias_table , overwrite = TRUE)

# * Check against internal gene panels dataset ----------------------------------
# x <- gnomeR::gene_panels %>%
#   unnest(cols = genes_in_panel)
#
# im <- x %>% filter(gene_panel %in%
#                      c("IMPACT230", "IMPACT279", "IMPACT300", "IMPACT341", "IMPACT410", "IMPACT468", "IMPACT505"))
#
# d <- setdiff(im$genes_in_panel, impact_alias_table$hugo_symbol)
#
# al <- map(d, ~get_alias(.x))
#
# names(al) <- d
#
# # should be zero
# al <- compact(al)

# GENIE ------------------------------------------------------------------


# * Get GENIE genes -----------------------------------------------------------

library(genieBPC)

genie_panels <- genieBPC::genie_panels

# Replace MSK-IMPACT* with IMPACT* because these tend to have the newer aliases
genie_panels <- genie_panels %>%
  mutate(Sequence.Assay.ID = str_remove_all(Sequence.Assay.ID, "MSK-"))

setdiff(genie_panels$Sequence.Assay.ID, gnomeR::gene_panels$gene_panel)

genie_genes <- gnomeR::gene_panels  %>%
  filter(gene_panel %in% genie_panels$Sequence.Assay.ID | gene_panel == "IMPACT505")

# unnested long version
genie_genes <- genie_genes %>%
  unnest(everything(),
         keep_empty = TRUE) %>%
  rename("gene" = genes_in_panel,
         "entrez_id" = entrez_ids_in_panel)


# * Get Aliases  --------------------------------------------------------

# get all aliases for impact genes
genie_alias <- genie_genes %>%
  select(gene) %>%
  dplyr::mutate(alias = purrr::map(gene,
                                   ~tryCatch(get_alias(.x),
                                             error = function(e) NA_character_)))

genie_alias$alias <- map(genie_alias$alias, ~as_tibble(.x))

# unnested long version
genie_alias_table <- genie_alias %>%
  unnest(everything(), keep_empty = TRUE) %>%
  rename("alias" = value)

genie_alias_table %>%
  filter(gene == alias)

# gets rid of NAs (genes with no aliases)
genie_alias_table <- genie_alias_table %>%
  filter(gene != alias)

# This may not be needed anymore, but leave for now
genie_alias_table$gene[map_lgl(genie_alias_table$gene, ~.x %in%
                                  genie_alias_table$alias)] %>%
  unique()

genie_alias_table <- genie_alias_table %>%

  # I don't think these are aliases even though they came up. They usually only come up one way
  # I think they are separate genes although I am not sure
  filter(!(gene == "SOS1" & alias == "HGF")) %>%
  filter(!(gene == "HRAS" & alias == "KRAS")) %>%
  filter(!(gene == "TP53BP1" & alias == "TP53")) %>%
  filter(!(gene == "EZH2" & alias == "EZH1")) %>%

  # are MLL2 and MLL4 interchangeable?! - JJ email on 1/25/2021 says KMT2B = MLL4
  filter(!(gene == "KMT2D" & alias == "MLL4")) %>%
  filter(!(gene == "KMT2B" & alias == "MLL2")) %>%
  filter(!(gene == "ATRX" & alias == "RAD54L")) %>%
  distinct()

# check we've recoded all
genie_alias_table$gene[map_lgl(genie_alias_table$gene, ~.x %in% genie_alias_table$alias)] %>%
  unique()

genie_alias_table <- genie_alias_table %>%
  rename("hugo_symbol" = gene)

genie_genes_clean <- select(genie_genes, gene, entrez_id) %>%
  distinct()

genie_alias_table <- genie_alias_table %>%
  left_join(., genie_genes_clean,
            by = c("hugo_symbol" = "gene")) %>%
  left_join(., select(all_genes, hugo_symbol, entrez_gene_id),
            by = c("alias" = "hugo_symbol")) %>%
  rename(alias_entrez_id = entrez_gene_id)

genie_alias_table <- genie_alias_table %>%
  distinct()

setdiff(genie_alias_table$hugo_symbol, impact_alias_table$hugo_symbol)
setdiff( impact_alias_table$hugo_symbol, genie_alias_table$hugo_symbol)

usethis::use_data(genie_alias_table , overwrite = TRUE)