Suppose I have a new gene assigned to a gene family. I also have a precomputed gene tree for this family. Is there a method to add the gene in the tree without having to compute a new MSA and avoiding to perform de novo tree reconstruction. I know that some multiple sequence aligners can add a new sequence to a precomputed MSA (ie MAFFT...). But what about tree reconstruction ?
Without rebuilding the tree from the alignment, how would you know where to insert the new gene ? Also how you would add the new gene to an existing tree would depend on how the tree was constructed i.e. maximum likelihood vs parsimony.
Treebest as Jean-Karim (hello!) said. Two possible ways. First, you can use neighbor-joining:
treebest nj -c known.nh aln.mfa
Treebest will not change the topology of known.nh even if the alignment strongly suggest a different one. Second, you can use an old Phyml:
treebest phyml -C known.nh aln.mfa
With the phyml command, you may also consider to change option -p as it may change the known tree topology when the alignment is strongly against it.
If you have the distance matrix from which the tree was built, you simply have to do a pairwise alignment between your new sequence and each of the others; that new vector of pairwise distances will become the new column in your distance matrix, then you can rebuild the tree. If you don't have the distance matrix, then you'll have to redo the MSA.
Without rebuilding the tree from the alignment, how would you know where to insert the new gene ? Also how you would add the new gene to an existing tree would depend on how the tree was constructed i.e. maximum likelihood vs parsimony.
treebest can constrain the tree to the one provided with the -c option
Thanks. Indeed TreeBeST is taking a species tree and can infer the orthologs and within-species paralogs.
You are right. But what I expect is to obtain a tree using the computed tree as guide. But perhaps this has no sense!