BA.5 "recombinant": evidence for it being artifactual

[hyanwong]

Of the recombination nodes affecting large number of samples, after #547 (with post-hoc fix in jeromekelleher/sc2ts#536), the only remaining "robust" recombination node in the ARG that is not a known pango X is the one associated with BA.5, node 1189192 (see #672 for another non-robust node). RE node 1189192 has the majority of the genome coming from the right parent (BA.4), up to position 26858, and the small remainder of the genome from a BA.2 parent. Here's the copying pattern

We believe this to be artifactual, because when we take a causal sample (e.g. SRR21340083 and match it into the ARG that we created just before the recombinant (the day before 2022-03-01), allowing intermediate nodes to be created in the manner of #536, we find that it is more parsimonious to match it in as a non-recombinant (the total cost goes from 8 to 7, presumably because we can swap out the recombination for an additional 3 de novo mutations). However, we can't do this post-hoc on the final ARG, because of subsequent tree-building. I haven't checked where the HMM matches the BA.5 sample, but I believe it to be onto a BA.4 lineage. It may be that BA.5 should have been seeded.

I have opened this up as an issue to track any comments about this remaining influential node, but we could also convert it to a discussion

[jeromekelleher]

I changed this to a Discussion as it is not something that can be closed by taking some actions, and is rather an open ended discussion.