Differences between GeneFull_Ex50pAS, GeneFull_ExonOverIntron, and GeneFull?

[cnk113]

Hey Alex,

I'm assuming your probably going to write something up on the changes, but from my searching in the docs it seems there is some prioritization of the reads when considered for unspliced counts? I guess what I'm asking is a bit more details on how the three differ?

Thanks,
Chang

[alexdobin]

Hi Chang,

right, I need to update the documentation for these options. Briefly:
GeneFull: simply considers overlap of an alignment with the full gene locus (including exons and introns).

GeneFull_ExonOverIntron: prioritizes purely exonic alignment (same as Gene option) over overlap with the full gene locus.
This is important, for instance, for a read that overlap gene A exons and gene B introns (i.e. gene A exons are located within gene B introns). With GeneFull option, such a read will be ambiguous and not counted. With GeneFull_ExonOverIntron it will be counted towards gene A.

GeneFull_Ex50pAS: more complex prioritization scheme. This options is most similar to CellRanger's intronic count option.
Then prioritization order is:

1. Exonic=100% (purely exonic as Gene)
2. Exonic=100% antisense (not counted)
3. Exonic>50% (i.e. >50% of aligned bases are exonic)
4. Exonic>50% antisense (not counted)
5. Intronic>50% (i.e. >50% of aligned bases are intronic)
6. Intronic>50% antisense (not counted)

The genes with the highest priority win. Only if a single gene has the highest priority (not antisense), it will be considered towards unique-gene counts of that gene.

Cheers
Alex

[joycekang]

Hi Alex,

Two quick clarifications:

1. Does the GeneFull_Ex50pAS option work with multimapping (EM) as well?

2. If we include two different versions of a genes (e.g. two different alleles for an HLA gene) as two additional chromosomes, labeled in the gtf file as the same gene (same gene_id and gene_name), will GeneFull_Ex50pAS still work? From a correspondence we had in December '21, you had mentioned that GeneFull might not work properly if the same gene is located in two genomic loci - is this still a concern?

Thanks so much for your hard work on this awesome tool!
Joyce

[cnk113]

I'm not Alex, but I can confirm 1 is working as that is what I've been mainly using.

[alexdobin]

Hi Joyce

1. I have not seriously tested it, but Chang Kim (see above) did - thanks!

2. GeneFull_Ex50pAS should indeed work properly for gene that have transcripts in different genomic loci, even different chromosomes.

[ColeWunderlich]

Hey Alex,
Thanks for the clarification on the new options. I am still a little unclear on GeneFull though.

1. Does it count antisense alignments to the annotations?
2. How much of an overlap with an annotation is required for an alignment to be counted?
3. Are these rules the same for applying GX/N and gx/n tags? (if only GeneFull is used)

[alexdobin]

Hi Cole,

1. GeneFull does not count antisense alignments.
2. 1-base overlap with the gene locus (gene start - gene end) is needed to assign an alignment to a gene. Note that if a gene has transcripts that do not overlap each other, GeneFull will not work correctly, while GeneFull_Ex50pAS will work correctly.
3. Yes, these rules work for unique-gene GX/GN and multi-gene gx/gn tags.

[ColeWunderlich]

Hey Alex,

I apologize for taking so long to get back to you. Thank you for the clarifications they definitely make sense.

I wasn't aware there was an issue with GeneFull. What happens when there are non-overlapping transcripts? (or is there somewhere I could read more about this)

I've also noticed that GeneFull_Ex50pAS runs noticeably faster and writes less tmp information to disk than GeneFull, is that somehow related?

[alexdobin]

Hi Cole,

for non-overlapping transcripts, the full region between the left-most transcript start and the right-most transcripts end will be used on counting reads for the gene. To avoid this, you can assign different gene_id to each non-overlapping transcripts, and aggregate the counts in post-processing.

[pandh0607]

Hey Alex,
I noticed that you said，Note that if a gene has transcripts that do not overlap each other, GeneFull will not work correctly,
Why is that?

[alexdobin]

Hi @pandh0607

GeneFull assumes the gene as the region between the left-most gene transcript start and the right-most gene transcript end.
This works fine with standard annotations such as ENSEMBL/GENCODE.
However, if transcripts do not fully cover the gene region defined above, the reads that fall into this inter-transcript area will also be counted towards the gene.

GeneFull_Ex50pAS option only counts reads that overlap transcripts.

[pandh0607]

hi Alex,
Thank you very much for your reply，I still have some questions。
For the same gene, If a read is mapped to both exonic and intronic ,intronic and intergenic, who has the highest priority？