UCSC hg19 gtf (genePredToGtf OS incompatibility)

Question

[umn_bist]

So I found that the annotated GTF file for hg19 from UCSC table does not adhere to the standard GTF format. Thus, I've been getting a fatal error in STAR:

Fatal INPUT FILE error, no valid exon lines in the GTF file: /work/cellbiology/s167125/Documents/ucsc_hg19/ucsc.hg19.gtf
Solution: check the formatting of the GTF file. Most likely cause is the difference in chromosome naming between GTF and FASTA file.

I know that I can retrieve a good GTF file via genePredToGtf application but this is only compatible with Linux 64. I only have access to a Mac. I am wondering if there is an alternative method to retrieve a GTF for UCSC's hg19 reference genome.

Thank you for the help

[Devon Ryan]

Is there a reason you want to use the UCSC annotation? The one from Ensembl/Gencode is almost always better (there's a reason that UCSC now uses the copy from gencode).

[umn_bist]

Yes, so I checked the header of my refgenome (ucsc_hg19.fa) as well as its annotated gtf file (ucsc_hg19.gtf) and it uses 'chr' notation.

Digging further, I realized UCSC does not keep a GTF file of its gene structures - they are all in GenePred Format.

[Vivek]

You can export the UCSC gene predictions in GTF from the table browser.

[umn_bist]

That is what I thought as well, but see this wiki page

UCSC does not keep gene structures in GTF format, we use a single line format for a single gene with all the information about that gene in the single line: GenePred format.

Extracting GTF format files from the genePred format can be performed with the genePredToGtf: kent command utility.

At this time, this genePredToGtf command can provide better GTF files than available from the table browser.

[umn_bist]

To be honest, no. It's just something I had on hand and had generated the index using STAR already. I found that Alex Dobin of STAR recommends using genecode.

[Devon Ryan]

Yup, Gencode/Ensembl (they're more or less identical) are what you'll find most people (myself included) recommending.

[epigene]

@Devon Ryan, Could you say a bit more about why Ensemble annotation is better than UCSC's? Thanks!

[Devon Ryan]

It's more likely to represent the transcripts you see in your experiments.

[epigene]

@Devon Ryan, because Ensembl people curate the annotation better?

[Devon Ryan]

Ensembl and UCSC use completely different methods to arrive at the annotations (historically, at least for recent mouse and human annotations they should be the same).

[Vivek]

It says the most likely issue is the chromosome naming convention. So it could be as simple as adding or removing a "chr" from the GTF or reference file.

Answer 1

[umn_bist]

Deleted. See comment above.