Open jmonlong opened 5 years ago
I was thinking that we might want to pick examples that are more interesting visually, but where we think vg is doing a good job. The potential medical relevance of these doesn't really come through in these visualizations.
I'm a little concerned about the second example. It shows a very large SV candidate. Is that real? It looks to me like it might be an artifact of the rendering model.
It won't be easy to add reference path positions here, but we can do so manually for the start and end of the SVs. The problem is that the rendering system only sees the subgraph around the SV and read collection embedded in it.
For the ultimate publication, I would copy take the images directly from the presentation (right click in evince), rather than using screenshots. This will prevent image quality degradation. I've attached the pair of these images here.
Sure, I first selected exonic SVs but it's not really important here. Also I realized that the whole region in these two examples are within UTRs so adding a gene annotation would be a bit useless, just an horizontal bar spanning the entire graph...
I'll look for other examples. What kind of features do you think would be better to show? Maybe heterozygous variants would be more interesting.
(Thanks for the evince trick)
In the end, we might first go for the examples we chose originally, and maybe focus on the deletion.
So @ekg if you want to explore different parameters for the visualization, please look at the following deletion chr2:100282798-100282848
in the HGSVC dataset.
Maybe zooming a little bit, for example +/- 2x read length instead of +/- 5x read length, e.g. showing the chr2:100282400-100283200
region.
And if there is a way to make the graph part larger vertically (or the reads thinner).
Thanks!
Edit: Fixed the coordinates
Here are the screenshots from IGV.
I agree, if we can zoom in on the odgi ones by about twfold to better see the variants, that would be good. I think we can line them up with IGV pretty well just by scaling one or the other. If there's no text, warping shouldn't show.
I tried to align the two figures and increase the graph topology's "height". We could use that for the submission or maybe it will be prettier to remake the figure directly with the zoom (and a bigger graph if possible/thinner reads). What do you think @ekg?
I changed fig3 in #100 to the version above using an updated zoomed-in version that Erik provided. It looks like this.
I added two examples, from @ekg 's slide deck, as fig 3: an insertion and a deletion found by vg only in HGSVC.
chr12:116277336-116277336
(in short tandem repeat region though)chr2:100282798-100282848
We could try to polish them by maybe:
Any other ideas?