MRCA for a set of taxa

A user just asked:

Do you know any function that could allow finding the deepest node ancestral to a group of tips? (I mean the oldest MRCA of a group of tips.)

The ape function mrca computes the pairwise most recent common ancestors for all the tips in a tree; however how do we take a set of pairwise MRCAs and identify the one that is also the MRCA of the whole group?

Well, by convention in "phylo" objects, node numbers increase along any path from the node. Logically, therefore, we ought to be able to pick the pairwise MRCA with the lowest number as the MRCA of a list. Since "phylo" objects are somewhat flexible, I decided to be careful and construct a function that pairs the ape function mrca with the phytools function nodeHeights to find the node in a set of pairwise MRCAs that is closest to the root node (i.e., it has the lowest height). Also, logically, this should be the MRCA of a set of tips.

The function looks as follows:

oldest.mrca<-function(tree,tips){
   H<-nodeHeights(tree)
   X<-mrca(tree)
   n<-length(tips)
   nodes<-height<-vector(); k<-1
   for(i in 1:(n-1)) for(j in (i+1):n){
      nodes[k]<-X[tips[i],tips[j]]
      height[k]<-H[match(nodes[k],tree$edge[,1]),1]
      k<-k+1
   }
   z<-match(min(height),height)
   return(nodes[z])
}

Let's load the function & phytools and then see how it works. First on an ultrametric tree (simulated using phytools pbtree):

> tree<-pbtree(n=20)
> plotTree(tree,node.numbers=T,pts=F)

> oldest.mrca(tree,c("t14","t18")) # two taxa case
[1] 36
> oldest.mrca(tree,c("t1","t3","t19"))
[1] 22
> oldest.mrca(tree,c("t2","t3","t9","t11"))
[1] 24
> oldest.mrca(tree,c("t2","t3","t4"))
[1] 21

Now, let's try the same with a non-ultrametric tree:

> tree<-rtree(20)
> plotTree(tree,node.numbers=T,pts=F)

> oldest.mrca(tree,c("t2","t12"))
[1] 30
> oldest.mrca(tree,c("t14","t19","t20"))
[1] 21
> oldest.mrca(tree,c("t10","t13","t20"))
[1] 25

Seems to work. . . .