Tumor phylogenetic tree software

Question

[ceruleanivy]

I would like to know if anyone has tried any software that constructs tumor evolution trees where the trunks represent the common mutations and the private alterations are noted on each branch. I can do it manually but the sample size is too big. Here's a good example here on Figure4 http://www.nature.com/nm/journal/v21/n7/full/nm.3886.html

Answer 1

[poisonAlien]

There are several ways you can do this.

1. By considering a mutation as present or absent (1,0) across all subsections of tumor.
2. By considering variant allele frequencies instead of binary 1s and 0s.

If doing by first way:

Phylip Character parsimony,

Phangorn r package

If doing it by second (more sophisticated) way:

lichee

Other related Biostar posts

R phangorn phylogenetic analysis from somatic mutations' binary table

Phylogeny based on presence/absence of genes

[ceruleanivy]

I only have metastatic sites and one sample from every primary tumor but that doesn't change anything, right ?

[poisonAlien]

So you just have two sections (Primary and metastatic) from each sample ?

[ceruleanivy]

No, I have only one primary and one to four metastatic sites, varying among patients.

[poisonAlien]

Yeah, it dosn't change much. However more samples is always better.

Answer 2

[Noushin N]

To complement the previous poster's answer, the first approach usually results in a tree phylogeny of tumor samples; while the second approach models the clonal composition of tumor samples.

In the case of second approach, you may find SCHISM (SubClone Hierarchy Inference from Somatic Mutations) of interest. SCHISM source code and tutorials are available on github.

[ceruleanivy]

What would you choose in order to construct something like Figure 4a in the Nature paper that I cite on the first post?

[Noushin N]

The cited figure looks like a subclone phylogeny.

[ceruleanivy]

One last thing; do you know if SCHISM is suitable for single nucleotide variations? I'm asking because the tutorials on wiki only show instructions for CNAs.

[Noushin N]

Yes. Single base substitutions and indels are the primary mutation classes used in SCHISM. I am not sure where in the wiki you are referring to.

[ceruleanivy]

It's the fifth column on the tutorial(input data) where it says copy number. I only have SNP data from my patients.

[Noushin N]

Right. The program requires CN to estimate mutation cellularities (cellular prevalence) from allele frequency; but the phylogeny is on the level of single base substitutions or indels; not CN events.

[ceruleanivy]

So, what am I supposed to enter in the fifth column ? The sequencer output provides (among others) reference and variant allele with their frequencies/coverage, mutation IDs, tumor cell content (%). My current database is nearly 4000 lines.