Specifying targets in a bed file

[looselab]

Specifying targets in a bed file

[alexomics]

This is something that we want to incorporate. It will definitely make downstream analysis easier!

[wdecoster]

So right now that's not yet possible?
Then we will put our targets in the toml format in an array, and we then use coordinates in chr1:123456-234567 format?

[mattloose]

It's almost possible... You can give a file as the parameter for a target:

e.g.

targets = "/path/to/your/targets/cancer_panel_and_MHC_targets.txt"

And that file will be of the following form:

chrX,71234624,71259140,+
chr19,53516024,53585269,+
chr12,6661648,6694510,+
chr19,21500564,21543078,+
chr7,57114614,57144864,+
chr18,25056926,25357152,+
chr22,28878592,29062487,+
chrX,15785472,15828260,+
chr6,28452797,33473354,+
chr6,28452797,33473354,-

So it isn't yet (quite) a bed file - but its close. Note the strand information. - so it's chr,start,stop,strand

[alexomics]

So currently, the targets are only in the formats:

chr1

or

chr1,0,1000,+

More generally it should be either an entire record, so just the contig name or coordinates on a record so: contig,start,stop,strand

These should fit the patterns:

"^[^,]+$"
"^.+,[0-9]+,[0-9]+,[+-]$"

[mattloose]

Like what Alex said ;-) - But noting that you can pass this in as a file.

[wdecoster]

Alright, we can work with that! Thanks!

You have me confused here for a bit that we need to specify the strand. Won't the alignment consider the reverse complement by default?

[mattloose]

The aligner does, but this is to capture the case where you are targetting say a 10 kb region with reads of mean length 10 kb. You might want to offset your target regions such that you capture any read which starts within 5kb upstream of your target. So your target might be:

chr1:10000-20000

But you might target:

chr1,5000,20000,+
chr1,10000,25000,-

This would capture reads that have some probability of extending in to your region of interest.

Does that make sense?

It may be over-engineered right now which is why we are going to make a bed format compatible (with possibly automatic self correcting strand offsettting - PASCSO?! down the line!).

[wdecoster]

Ah gotcha. I already padded the intervals to take those flanking-potentially-on-target reads into account, but indeed not strand specific.

[alexomics]

@wdecoster I've added some description to the TOML.md file. Does this make things clearer?

[wdecoster]

Yep, all clear now!

[ythuang0522]

@mattloose We are still not quite sure of this bed formatting. Our interpretation is, for any given target gene, e.g., chrX,71234624,71259140, we must duplicate two coordinates in the target file even without extension.
i.e.,

chrX,71234624,71259140,+
chrX,71234624,71259140, -.

Without this duplication we will miss half alignments on one strand. Is that true?

[alexomics]

@ythuang0522 Yes, to cover both strands you must specify both + and -.

[ythuang0522]

Thank you.