biomaRt - snoRNA gene ids identifying as protein coding gene that map to same locus

Question

[kvince.888]

I'm attempting to first generate a data frame of only homo sapiens protein-coding genes using the following code:

library(biomaRt)
ensembl=useMart("ensembl")
ensembl=useDataset("hsapiens_gene_ensembl",mart=ensembl)
filterlist <- list("protein_coding")
protcoding.genes.with.id=getBM(attributes = c("hgnc_symbol","entrezgene", "external_gene_id"),filters = c("biotype"),values = filterlist, mart = ensembl)

I get a data frame with 22710 observations. But I've found that snoRNA entrez gene ids are also included as the protein coding gene that they share a locus with. For example:

quest<-protcoding.genes.with.id[grep(paste("^","GNL3",sep=""), protcoding.genes.with.id$external_gene_id, ignore.case=TRUE),]

View(quest)

    row.names    hgnc_symbol    entrezgene    external_gene_id
1    12364    GNL3L    54552    GNL3L
2    15208    GNL3    100113381    GNL3
3    15209    GNL3    26354    GNL3

Row 2 with entrez gene id 100113381 is actually SNORD19B...which maps within one of GNL3's introns.

How can I get only the actual protein coding genes and no snoRNAs?

Thanks!

[Neilfws]

Interesting; returning the attribute gene_biotype gives values of protein_coding for all the results. Maybe it's a quirk of how gene IDs are mapped? I'm sure Emily can explain.