Trees output full results

[underbais]

Hello

In PhyloWGS we get mut assignments and trees structures (e.d.summ.json) to play with:

My question is do we get just one perfect tree in Pairtree as a final result or there is away to look at all of them? If so, how? And how do I get mut assignments and tree structure out of it?

Thanks Chingiz

[jwintersinger]

Hi Chingiz,

I should document this in the README -- thanks for reaching out! Pairtree produces multiple posterior samples, just like PhyloWGS.

• To summarize the posterior, run bin/summposterior (pass --help for usage information). This will produce a graph structure showing the different possible evolutionary relationships implied by the posterior samples, and also shows the best 20 trees.
• If you want to delve into more detail for any one of those trees, run bin/plottree with the --tree-index <integer> option. The index you pass here corresponds to the indices displayed by summposterior in the top 20 trees.

Please let me know if you have any other questions!

[underbais]

Hi Jeff,

Thanks a lot, very helpful. I guess my issue is with results.npz format. In PhyloWGS mutass.zip has all mut assignments and summ.json has all tree structures. Just wonder how to extract those from npz. More info on the npz output structure would really help.

Best, Chingiz

[jwintersinger]

Hi Chingiz,

No problem! Unlike PhyloWGS, Pairtree uses the same SSM clustering for every tree it samples. This means that the assignment of SSMs to subclones is just determined by the variant clustering you provide in the params.json input file. I.e., the mutations for tree node 1 are those in the first cluster, the mutations for node 2 are those in the second cluster, and so on.

To get results that are easier to parse, run bin/plottree with the --tree-json <name>.json option. That will give you output like this: https://gist.github.com/jwintersinger/5670966f731204ffc558bc04df7ff403.

• phi gives the subclonal frequencies, ordered by each sample.
• eta gives the subpopulation frequencies, ordered by each sample.
• parents gives the tree structure. parents[i] gives the index of the parent for tree node i+1. So, the example above has the parent of node 1 as 0, the parent of node 2 as 1, and the parent of node 3 as 0.

The NPZ file will have that information for every tree, but if you're just trying to parse the information for one tree, the JSON should be easier to work with. I'll be happy to give more details if you have other questions!

[oghzzang]

Hi, I have a question related to this thread. Then, which mutations does the "node0" have?

[ethanumn]

Hi Hayley -

Node 0 represents the non-cancerous genome. It does not contain any of the mutations that you might have included in your analysis. We've written extensively about how to interpret the trees output by Pairtree here and here.

Ethan

[oghzzang]

Many thanks!!!