Why many ENSEMBL gene ID cannot be converted to Entrez ID using org.Hs.eg.db in R?

Question

[biock]

Hello!
I'm analyzing some RNA-seq data with edgeR. According edgeR manual, we can use org.Hs.egENSEMBL database in org.Hs.eg.db package (version: 3.7.0) to convert ENSEMBL gene ID (ENSGxxxxxx) to Entrez ID. However, I found there are many ENSEMBL gene IDs cannot be found in egENSEMBL database. There are 30292 ENSEMBL ID records in egENSEMBL, while there are 58721 ENSEMBL gene IDs stored in GENCODE GRCh38 annotation file. Should I exclude genes being not in egENSEMBL database for downstream differential expression analysis just as the edgeR manual do?

Thank you!

Codes in edgeR manual (I use egENSEMBL instead of egREFSEQ in my pipeline):

# y is DGEList object
idfound <- y$genes$RefSeqID %in% mappedRkeys(org.Hs.egREFSEQ)
y <- y[idfound,]

Answer 1

[Jean-Karim Heriche]

Ensembl and RefSeq are fundamentally different resources. Ensembl is an annotation of a reference genome whereas RefSeq is a collection of sequences with annotations. They differ in particular with respect to how they define a gene. In Ensembl, a gene is an annotated locus on the reference assembly. In Refseq, it seems to be an extra attribute assigned to transcript sequences. How RefSeq assigns genes to sequences has never been clear to me. In general, I wouldn't recommend mixing references, i.e. either work with Ensembl or work with RefSeq.