RegionGrowth documentation

[baozg]

Hi, @heringerp

I found in 0.5.0, you mentioned that including the RegionGrowth. But it seems not present in the documentation yet.
Is it an new feature not relased yet?

[heringerp]

Thanks for the pointer @baozg!

Sorry, I haven't added the documentation for it yet. I've just added !RegionalGrowth, but I still need to add the other pages. Also the documentation still is a bit rough.

The features however are already available to use in panacus >= v0.5.0

[baozg]

Thanks for quick update. That's exactly the feature I am looking for. Happy to try in our pangenome graph!!

[baozg]

Is it possible to set an bed (not equal window size) for ReionalGrowth and the normalized by ref or median hap length for growth analysis?

[heringerp]

Unfortunately not yet, but in principle this should be possible to implement. At least for the ref length, median hap length not so much as panacus does not collect this info from the GFA file.

[baozg]

Ideally, given one region with all node and walk/path information, pancanus should able to know this? Sometimes, ref just the outlier in the population, but even that may still good.

[heringerp]

Yes, in theory it should be. However, currently panacus does not store the walk/path information as an explicit sequence but instead only saves which nodes/edges it touched. From that we cannot know how long a path really is as it might hit the same node multiple times, while panacus only counts that as hitting the node.

[baozg]

How panacus deal with clipped sequence in the MC graph? Is those empty window was caused by clipping?

[heringerp]

Empty windows in the Regional Growth analysis are caused by panacus not being able to predict the growth when there is no good fit for the growth curve (or technically Fnew). Clipped regions in an MC graph only contain a single sequence (namely the reference) and therefore there is no growth and thus no values there. So the clipped regions always show up as empty windows, but there are some empty windows that are caused by other things than clipping

[baozg]

Thanks for quick reply. I will dig these more and feedback later if I could confirm it comes from MC or Panacus. Run with a small subet with pggb and MC may could find out who contribute to empty more. Sliding window somes may break around some large SVs

[heringerp]

Hi @baozg

I'm currently working on the bed-file support for custom windows. Do you need this to also allow over-lapping windows or are non-overlapping windows fine?

[baozg]

I think overlapping window may more make sense in case some large SVs exactly around the window boundary. But if this need more work, non-overlapping windows is enough.

[heringerp]

Thank you! I will write it as to allow overlapping windows (probably it will get slow if too many windows overlap in a single place or for nested windows, but this shouldn't be a problem in that case)