Ideas for viz

[hyanwong]

I just had a good idea for making sense of the ancestry viz, so I'm jotting it down here. Instead of trying to visualise everyone, we could focus on just one leaf node (e.g. a Bougainville), and plot lines joining the sample to our geographical position of its ancestors. For each point in the genome there will therefore only be a single line, tracing a path back to the grand MRCA in Africa. If we want, the opacity (alpha) value of the line could reflect the span of the genome covered by that particular route.

In the case of Bougainville, we are interested in which lines "pass through" Denisovan-like ancestry. As we discussed, I'm keen that this doesn't just involve direct descent from the proxy ancestors, but also descent from Denisovan-like individuals. My suggestion is to imagine colouring in the Denisovan samples in (say) orange, and imagine that colour gradually changing to the default (blue?) as we go back in time - the exact mapping of colour to time that works vest for the viz would need to be determined by trial and error. We needn't actually plot the Denisovan ancestors, merely assign a value to each lineage, which is "degree of closeness to the Denisovan sample(s). We then plot the Bougainville lineages using the colour obtained from the first time that lineage joins with a Denisovan lineage.

This doesn't require us to set any specific parameters about "how Denisovan" an ancestor is - we just use time as a continuous colour scale. It could also be extend it to a few basically identical Bougainville samples.

I guess the key here is to simplify the viz by focussing on lineages that go back from only a single genome at a time.