How do you process aliases of gene names/gene IDs and obsolete gene names?

[lokapal]

Hello, geneslator developers!

As you correctly pointed out, the problem of association geneID and symbol name is very complex. Do you use in your database HGNC gene names aliases databases? 

EMBL list:
https://ftp.ebi.ac.uk/pub/databases/genenames/hgnc/tsv/hgnc_complete_set.txt
and NCBI list:
ftp://ftp.ncbi.nlm.nih.gov/gene/DATA/GENE_INFO/Mammalia/Homo_sapiens.gene_info.gz

The problem with the gene names in human genome is even more serious: after EMBL annotation v.29 (if I remember correctly) a lot of symbol names of IDs were removed, so if we need to compile the most full database of gene names and IDs it should be something like to compiling a dictionary

GENETABLE <- as.data.frame(read.table("names.v112.txt", sep = "\t", header = T, stringsAsFactors = FALSE, row.names = NULL, comment.char="", quote=""))
GENETABLE <- rename(GENETABLE, GeneID = gene_id)
GENESV29  <- as.data.frame(read.table("names.v29.txt", sep = "\t", header = T, stringsAsFactors = FALSE, row.names = NULL, comment.char="", quote=""))
GENESV29  <- GENESV29 %>% dplyr::select(GeneID, gene_name)
# reading all aliases from HGNC converted database
ALIASES   <- as.data.frame(read.table("Homo_sapiens.converted", sep = "\t", header = T, stringsAsFactors = FALSE, row.names = NULL, comment.char="", quote=""))
ALIASES   <- rename(ALIASES, gene_name = GeneName)

# here we're binding the full list. All source DFs must have the columns with the same names
DICT <- dplyr::bind_rows(list(GENETABLE, GENESV29, ALIASES)) %>% distinct()

And then use function like

names2IDs <- function (dframe) {
  final <- dplyr::left_join(dframe, DICT, by = c("Genes" = "gene_name"),multiple = "all") %>%
           dplyr::mutate (GeneID = coalesce(GeneID, Genes)) %>%
           dplyr::rename (gene_id = GeneID)
  return (final)
}

in general case to get all IDs that matches names list. Please, clarify your approach in the documentation - what kind of aliases do you employ, and how do you treat gene names that were discarded (but still present in many databases/publications)? How do you process cases when one geneID -> many gene names (what gene name do you select in that case?) and rare cases of gene copies when one gene name -> many geneIDs?
Probably, you can contact with genecards.org and get from them aliases database.
Also it would be excellent if you will provide in the vignette the example how to convert all gene IDs/names from the specific column of dataframe/data.table objects. Option "all" or "any" is very important for wet biologists and should be the option of matching function.

[GMicale]

Hello lokapal,
thanks for your interest in geneslator package.
Our databases have a column called SYMBOL with the official symbol of the gene and a separate ALIAS column that contain all possible other names. Aliases are taken from NCBI and Ensembl, which are the most reliable source of information for gene associated data. We have documented all the steps we perform for collecting and integrating aliases and all other gene data in the companion repository https://github.com/knowmics-lab/geneslator-data, where we host monthly updates of our annotation databases. To build our databases we also consider older versions of NCBI and Ensembl database (since 2015). This means, for example, that aliases of old genes or old aliases of current genes are fully integrated into our databases. Concerning multiple mappings, they are already handled by our mapping functions. The general rule is that, in case of multiple mappings between keys and values, all one-to-many mappings are reported. The alternative "any" option is provided only in the case of orthologs mapping, where the user can choose if getting only the first ortholog found for a gene or all possible orthologs of a gene.