Reference Gene Sequence file

[Jluis]

Dear all,

I've just arrived to the NGS field and I'm currently facing many issues with different bioinformatic analysis workflows. Maybe the questions I may ask look naive for most of you, but I hope you can help me even if the answer is obvious to you.

The issue I'm on right now is this:

Is there any public database or repository where I can download a file where all the genes of an organism (in this case those corresponding to the hg19 sequence version of the human genome) are displayed showing the name of each gene, its coordinates and the chromosome its located at?

The file format should be somehow equivalent to this fields:

#GeneID Start End Chromosome ...



Thanks in advance

[jimmybee]

You will be able (I assume although ive never worked with the human genome before) to download gff format files http://www.sanger.ac.uk/resources/so.../gff/spec.html) from the NCBI ftp database which has all the information that you need.

The ftp site is here: ftp://ftp.ncbi.nlm.nih.gov

[Richard Finney]

UCSC has good data.
Go to http://hgdownload.cse.ucsc.edu/goldenPath/hg19/database
and download refFlat.txt.gz

Direct link:

http://hgdownload.cse.ucsc.edu/goldenPath/hg19/database/refFlat.txt.gz

This is the refseq "definition" of the human genes from the NCBI at NIH.

It is carefully curated "by hand".

Other folks have their own "guesses" as to what the comolete set of human genes is.

It's a "moving target" and the final catalog of what genes are in the genome isn't done.

refseq is a pretty good picture, based on current understanding, of what's there.

[Jluis]

Dear jimmybee and Richard Finney,

The information you've given me here has been of great value. I've followed your instructions and downloaded the references. Now I have to see if I can establish correspondences between the info in those reference tables and the info I obtained from my bowtie alignment files against the hg19 genome. I'll keep you informed on how I'll try to do this.

Best wishes and thank you once more for your kind help.

[laura]

You can also get this data from Ensembl

ftp://ftp.ensembl.org/pub/current_gtf

The human data is mapped to GRCh37/hg19

ftp://ftp.ensembl.org/pub/current_gtf/homo_sapiens/

[Jluis]

Dear Laura,

Thank you very much for your feedback. It was last the piece of information I was looking for, in order to compare my mapping results to the table of counts I had generated using R , whose reference was the ensembl annotation of hg19.

You've all help me a lot with this issue.

[shyam_la]

Doesn't ensembl gene annotation have the problem of "too many transcripts"?? I tried using ensembl gene references with snpeff to annotate a list of variants I had generated (for a tumor-normal pair) and the output file had increased in number of lines by more than 5 fold. Multiple transcripts are (average of 3-4) assigned to the same position. Doesn't this make downstream analysis difficult?

I am new to Bioinformatics too. And this has been bugging me for a few days.

Is it okay to use refseq gene definitions, if my variants were obtained by aligning to GRCh37.67?

Thanks.

You can also get this data from Ensembl

ftp://ftp.ensembl.org/pub/current_gtf

The human data is mapped to GRCh37/hg19

ftp://ftp.ensembl.org/pub/current_gtf/homo_sapiens/

[laura]

An interesting complaint, I have rarely seen Ensembl being accused as having too many alt splice forms, normally people complain about too few

The majority of loci have more than one transcript which will be expressed at different times in different tissues

This does give you more information but this extra information can be important

The ensembl variant effect predictor script does have an option to output only the most severe consequence per gene and if this consequence occurs in more than one transcript the consequence is chosen arbitarily

http://www.ensembl.org/info/docs/variation/vep/vep_script.html#output

The best solution to this problem is of course to have expression data for tumor and normal tissue so you can actually know which transcripts from a particular loci are being expressed (using rnaseq or microarray data) so you only need consider the appropriate transcripts on the basis of the expression data

All the annotation data should be based on the same reference, for the autosomes (and mostly for X and Y) hg19 and the primary GRCh37 should be identical in terms of sequence and coordinates. UCSC choses to use its own naming convention for its assemblies rather than the official GRC name

[shyam_la]

All the annotation data should be based on the same reference, for the autosomes (and mostly for X and Y) hg19 and the primary GRCh37 should be identical in terms of sequence and coordinates. UCSC choses to use its own naming convention for its assemblies rather than the official GRC name

You mean the current hg19, if I download it today and GRCh37.67 will have the same sequence and coordinates??

[shyam_la]

An interesting complaint, I have rarely seen Ensembl being accused as having too many alt splice forms, normally people complain about too few

That is an interesting statement! Ensembl gene set has 57000 "genes" while RefSeq has only 24000. I would have expected that to be a non-rare complaint..

Kindly clarify.

[laura]

hg19 only represents the primary GRCh37 assembly as far as I am aware (someone please correct me if I am wrong) so long as you stick to the primary chromosomes (so 1-22, X and Y) the coordinates should be identical

[laura]

That is an interesting statement! Ensembl gene set has 57000 "genes" while RefSeq has only 24000. I would have expected that to be a non-rare complaint..

Kindly clarify.

The Refseq cdna set will probably only considers protein coding transcripts/cdnas

Ensembl's 57000 includes not just protein coding genes but also pseudogenes, ncRNAs and some other things

If you go to biomart and just look at protein coding genes you get 21976 genes

[shyam_la]

hg19 only represents the primary GRCh37 assembly as far as I am aware (someone please correct me if I am wrong) so long as you stick to the primary chromosomes (so 1-22, X and Y) the coordinates should be identical

Well, I have come across another source that said that hg19 is also updated but the name stays the same. There is no version number unlike GRCh37 and hence the latter should be preferred so that you can keep track of which version you are using.

You seem to suggest that hg19 has stayed unchanged since Feb 2009, when it was first released.

So many conflicting ideas on the internet! Sigh.. A newbie is bound to get lost..

[laura]

I am no expert in UCSC so I am just going by what the browser is called, If your other source is the UCSC help pages or something similar I would trust it

I will point out though that the primary assembly (ie the main chromosome 1-22 and chrX and Y) has not changed since GRCh37 was released in 2009 so if you are just using the main chromosomes without any repeat masking then it doesn't matter what the pN number is the chromosomes are identical

The version only matters if you are also considering alternative haplotypes and GRC fix patches

[shyam_la]

I am no expert in UCSC so I am just going by what the browser is called, If your other source is the UCSC help pages or something similar I would trust it

I will point out though that the primary assembly (ie the main chromosome 1-22 and chrX and Y) has not changed since GRCh37 was released in 2009 so if you are just using the main chromosomes without any repeat masking then it doesn't matter what the pN number is the chromosomes are identical

The version only matters if you are also considering alternative haplotypes and GRC fix patches

http://snpeff.sourceforge.net/faq.html#When_I_look_at_the_UCSC_borwser_for_hg19,_it_doesn%27t_match_the_information_from_snpEff?

This was my source..

But I know you are correct now!

Just one more question: Number of FIX Patch scaffolds aligned to the Primary Assembly 60
This is from the statistics page for GRCh37.p7. What does 60 mean here?