[Biopython-dev] Calculating bootstrap values from a set of trees

[Peter Cock]

On Mon, Sep 12, 2011 at 2:14 PM, Eric Talevich <eric.talevich at gmail.com> wrote:
> Hi Peter,
>
> On Mon, Sep 12, 2011 at 5:07 AM, Peter Cock <p.j.a.cock at googlemail.com>
> wrote:
>>
>> Hi Eric,
>>
>> I'm wondering if there is any code in Bio.Phylo for calculating
>> bootstrap values from a set of trees?
>>
>> e.g. I have a master tree created from an alignment, and 1000
>> bootstrap trees (created from 1000 re-sampled alignments). I want to
>> annotate each branch with the number/percentage of times is it found
>> in the 1000 bootsrap trees.
>
> I haven't implemented this in Bio.Phylo yet, unfortunately. The closest
> thing is Bio.Nexus.Trees.consensus. It would be worthwhile to port this to
> Bio.Phylo eventually.
>
>
>> I once implemented this in python using binary strings to represent
>> each branch as a split or partition of the nodes into two groups. I'm
>> not sure where I put this script... but it pre-dated Bio.Phylo anyway.
>>
>> Alternatively, which standalone tool would you recommend for this?
>>
>
> I think Phylip's seqboot and consense will do the trick.

http://evolution.genetics.washington.edu/phylip/doc/consense.html

My understanding was Phylip's consense takes a set of trees
and finds a consensus - there is no obvious way to tell it you
want to use a particular pre-determined tree.

>
> Normally I let RAxML do this sort of thing for me.
>

I'm unclear if RAxML will accept some 3rd party master tree
(via -t) and a set of bootstrapped trees (via -z) without also
wanting the original alignment and a choice of model...

My reason for wanting to decouple bootstrapping the trees
and applying the bootstraps to the master tree is for splitting
large jobs across a cluster. Each cluster node can generate
bootstrap trees independently of the other cluster nodes
(no network IO or synchronisation needed). These trees are
then collated (concatenated into a big multiple entry tree
file), with the final step combining the bootstrapped trees
onto the master tree to assess support being comparatively
quick.

Peter