Duplicated gene symbols in the output of msigdb_download

[kotliary]

Currently the msigdbr package includes Ensembl IDs in the output data frame of gene sets. Since there are multiple Ensembl IDs corresponding to some genes, your msigdb_download function returns duplicated genes in some gene sets. For example, check HALLMARK_APICAL_JUNCTION (Human, H category).

You just need to add distinct() %>% after the second line below in the db_msig.R (lines 148-151):

    mdf <- msigdbr(species, category, subcategory) %>%
           dplyr::select(gs_name, gene_symbol) %>%
           as.data.frame() %>%
           stats::aggregate(gene_symbol ~ gs_name, data=., c)

[kotliary]

By the way, the mdf data frame can be easily converted to the list with base split function:

mdf <- msigdbr(species, category, subcategory) %>%
    dplyr::select(gs_name, gene_symbol) %>%
    distinct()
gsets <- split(mdf$gene_symbol, mdf$gs_name)

Note that the gene set names will become list names automatically.

[anfederico]

Thanks, good catch! Will fix ASAP. Fortunately this bug ends up getting canceled out during the enrichment steps where both the signature and genesets are reduced to unique elements.

signature <- unique(signature)
genesets <- lapply(genesets, unique)

[kotliary]

Yes, it's true. The reduce method also removes duplicates.

[anfederico]

What version of msigdb are you using?

My version doesn't seem to have duplicated gene symbols in any of the geneset collections I've tested.

hypeR::msigdb_version()
#> [1] "v7.2.1"
g <- hypeR::msigdb_download("Homo sapiens", category="H")
table(duplicated(g$HALLMARK_APICAL_JUNCTION))
#> 
#> FALSE 
#>   200

^{Created on 2021-08-23 by the reprex package (v2.0.0)}

[kotliary]

I use 7.4.1. I believe they added Ensembl ID output in the latest version., around May this year.

[anfederico]

Fixed, thank you.