RefSeq and hg19, what's the relationship?

[kreitinger]

I have noticed that some RefSeq sequences downloaded from NCBI do not completely match the human reference genome hg19. For instance, the sequence of NM_000018.3 contains a C at nt position 224, but there is a T in the same position on the human reference genome (hg19).

The RefSeq manual describes its annotation process. RefSeq entries are created from INSDC & GenBank data. There is no mention about the human reference genome from GRC.

But, one can download a RefSeq GTF file and use the GTF coordinates along with the human reference genome to extract a synthetic transcriptome.

Does anyone understand the relationship between the RefSeq GTF file, the RefSeq sequences retrievable through NCBI, and the human reference genome?

Much thanks,

Gloria

[Neil.Miller]

Hi Gloria -- my understanding is that the RefSeq transcript sequences are derived experimentally and separate from the reference genome. This means that the sequences will not match the genome exactly -- the genomic positions for the transcripts are determined through a sequence alignment and will result in the occasional mismatch or even small in/del.

This has some unfortunate consequences -- basically, it means that if you pull the exon sequences from the genome based on the annotation and concatenate it together, you won't get the exact sequence listed in RefSeq. It also means that if you look at the translation frame notations for the edges of the CDS annotations, there are a number of invalid transitions (i.e. the last base of one exon might be listed as position 1 of the codon while the first base of the next exon is listed as position 3)

Hope this helps

--Neil

[kreitinger]

Hi Neil,

Thank you for clarifying the relationship. So the RefSeq GTF file is created by aligning the RefSeq sequences to the reference genome and then deriving the RefSeq gene coordinates.

Do you by chance know if this is the same situation for Ensembl, Gencode, other databases?

Gloria

[Neil.Miller]

There are probably others who can better comment, but there is a close (though not exact) correlation between the ENSEMBL transcripts and the RefSeq. I've heard that the ENSEMBL annotation has been curated a little more closely to minimize the mismatches between the transcript and genome sequences but haven't looked at it much myself to verify.

[GenoMax]

Annotation in Ensembl/VEGA comes from the HAVANA group at Sanger. http://www.sanger.ac.uk/research/pro...genome/havana/ Click on the "annotation" tab to get additional information.