feature annotation

[anoronh4]

I am currently using sjdbList.out.tab from a star_index run as a --reference to run portcullis. as it stands, i get coordinates but no annotation for the gene, transcript, or exon numbers. i am just wondering if there's any recommended process to annotate portcullis results, or if there's a way to tweak the input reference so that annotations are added.

the beginning of my --reference file is like this:

1	12058	12178	+
1	12228	12612	+
1	12698	12974	+

[anoronh4]

i tried out using junctools to create the input:

junctools convert -if gtf -of BED6 -o Homo_sapiens.GRCh38.111.junctools.bed6 ../results/references/GRCh38/gunzip_gtf/Homo_sapiens.GRCh38.111.gtf

as a result the first few lines of the result look like this:

track name="junctions"
1	12057	12178	ENST00000450305_junc1	0.000	+
1	12227	12612	ENST00000450305_junc2	0.000	+
1	12227	12612	ENST00000456328_junc1	0.000	+
1	12697	12974	ENST00000450305_junc3	0.000	+
1	12721	13220	ENST00000456328_junc2	0.000	+
1	13052	13220	ENST00000450305_junc4	0.000	+
1	13374	13452	ENST00000450305_junc5	0.000	+
1	15038	14695	ENST00000488147_junc9	0.000	-
1	15947	14969	ENST00000488147_junc8	0.000	-

from what i can tell none of the names in column 4 of the junctools result are included in the portcullis full run. so i don't think this a good solution.

[bmwx410]

I have noticed the reference not being utilised properly as well, seeing if I can code in a solution...

[bmwx410]

In case you are still looking @anoronh4 the tools works if you run each step individually. For some reason it doesn't read the -r reference flag in the portcullis full mode.
Also, your file needs to be in ebed format, junctools -of ebed for the GTF.
Hope this helps.

[anoronh4]

hi @bmwx410 , assuming you are one of the maintainers for the tool, do you think that this bug will be fixed in portcullis full any time soon? the note about ebed is also very helpful, thank you.

[bmwx410]

hi @anoronh4 I'm not actually, was just keen to make this tool work in my pipeline. I don't think this tool seems to be maintained at the minute. As far as I can see though doing the multistep is really not much slower than the full version, just a couple more lines of code.

[anoronh4]

@bmwx410 true, it's just that I'm using this tool with the nf-core nextflow framework. It's definitely easier to just upgrade the one-step module they already have wrapped than write a new multi-step module. Hopefully if you or I make an improvement they will approve it and publish.