GFF3

[Nate Hansen]

Hello,

I have been trying to run MAJIQ on a more recent human ensembl build but have bene running into issues with the conversion and making sure it is perfect for majiq.

I know you guys already have hg19 available for download since it was in one of the papers, but I was wondering if you also have hg38 available to be downloaded somewhere as this would make running majiq on the updated build much easier.

Thanks, 

Nate

[Nate Hansen]

Following up on this, I do not see a script available that converts either a gtf or gff3 from ensembl to the specific format required for MAJIQ. I am trying to run against GRCh38. Please advise, thank you!

[San Jewell]

Hi Nate,

In general you should be fine with using the gff3 that ensembl provides (i.e. https://ftp.ensembl.org/pub/release-114/gff3/homo_sapiens/Homo_sapiens.GRCh38.114.gff3.gz)

Some annotations do produce warnings like "Error, incorrect gff. exon doesn't have valid transcript" ; however, this does not indicate that the build will produce invalid results, rather, there exist a few exons which don't have a proper transcript specified in the annotation, so these exons will not be used downstream. For example, I see that the ensembl annotation has approximately 40 orphan exons. However, the rest of the annotation is usable and it should generally not be a cause for concern. Let me know if that answers your question.

Thanks,

-San

[Nate Hansen]

Hi San,

Thank you for the response. I have been using the gff3 that ensembl provides, but my bam files have a "chr" to label the chromosomes whereas the gff3 file does not contain that prefix and I am identifying 0 LSVs. Is that difference in the chromosome annotation the reason why no LSVs are identified?

Thanks,

Nate

[Nate Hansen]

Hi San,

Looks like that did seem to be the issue and when I matched those up everything ran smoothly.

Thanks,

Nate