I think that one thing that could be helpful, for example, is that I have a pairwise distance file with three tabs:

genome1  genome2  distance

And I have a perl script to total the distance although probably not optimized

perl -MList::Util=max -lane '
  $d{$F[0]}+=$F[2]; $d{$F[1]}+=$F[2]; # give distance 'points' to each genome
  $e{$d{$F[1]}}=$F[1]; $e{$d{$F[0]}}=$F[0]; # reverse-index which points belong to which genome
  END{
    $max=max(values(%d));   # find the max distance attributed to a genome
    print "$e{$max}\t$max"; # print the genome and its distance
  }
' < pairwise.tsv

However, I think that because of a sampling bias, it is not giving me the correct outgroup.