Open parlar opened 10 years ago
@dasmoth I was experimenting with this using a mapping chainset, and wanted to check with you what's the expected behavior for mapping with strand flip.
Here's a hypothetical illustration... Suppose there's a 3-exon gene on the reverse strand of chrZ.
If I do a mapping using a chainset that flips the strand in this region
chrZ 1000 1800 + chrZ 1000 1800 - 1 0 0 800
I believe Dalliance will give a lift-over in the form below. This is not really what we want for this use case.
An alternative interpretation of the liftover coordinates might give us something like this, which is visually what we want but with meaningless coordinates
I don't think either is actually what par wants, but just wanted to clarify with you what's the expected behavior with mapping segments with strand flip.
Well, let's be slightly more specific here: I can draw your test gene structure with the following line of BED data:
chrZ 1001 1800 test 1000 - 0 0 0,0,0 3 100,200,300 0,200,500
In Dalliance, this looks like:
...which I think matches what you expect. If I now map this using your suggested chain.bed fragment (I edited very slightly to make the target chromosome "chrZr" rather than "chrZ", to make it clear we're looking at the flipped version). Back in Dalliance, what I see is:
...which matches your second (green) mapped version and (unlike the red version) still makes biological sense.
How does this differ from what @parlar is looking for?
The one bit I don't understand is why you seem to have a mapping file which just flips the region around a single gene. Doesn't it make more sense to flip the whole chromosome? If this turns out to be a generally useful feature, we should then build some UI elements to indicate to the user that they're looking at a flipped view (the obvious one that springs to mind is that the genome ruler should probably display "real" chromosome coordinates -- descending from left to right -- rather than coordinate on the virtual flipped chromosome).
Or am I missing something here...?
No, I'm the one missing on stuff here :( I should have tested this more thoroughly, I apologize for that, and thanks for taking the time to reply.
The one bit I don't understand is why you seem to have a mapping file which just flips the region around a single gene.
That was a contrived example. Yes, the idea will be to flip every chromosome in full.
We should then build some UI elements to indicate to the user that they're looking at a flipped view
Yes, I'm mainly worried about the coordinates (both the ruler as well as the actual positions that show up if you click on a feature). If people are used to what reverse
does in UCSC which uses real coordinates, this might be misleading.
I guess it's possible to get the chromosome sizes (from the chainset or just store it inside the browser) and transform the imaginary coordinates in the ruler and feature pop-ups.... not sure if that's too hacky of an approach though. What do you think?
That is an important point regarding the info popups. Probably needs a flag somewhere to say "want unmapped coords in the info popups". I imagine that might also be important for #63.
To enable viewing of genes/exons in the forward order (5'->3') if present on the minus strand.
With reference to google groups discussion (https://groups.google.com/forum/#!topic/biodalliance-dev/OjYwCAk2XFo).
Kind regards,
Pär