I guess that you should consider your data as a binary matrix (presence/absence) and perform a cluster analysis accordingly.

There are many tutorials on how to do that in R (here and here).

I guess that when you talk about phylogeny you talk about a more complex situation. You have a sequence alignment and according with the method you use you need to reconstruct the probable evolutionary processes that are responsible for the divergence of the resequences. This is done considering different model of evolution and the frequencies with which a change happens to a specific position in your sequences. Then you need to create a tree like representation of the relationship amongst your sequences and also this can be done using different methods not only based on similarity onr dissimilarity values amongst your sequences.

Here there are a bit more details.

Try using PAST this can form neighbor joining tree from multivariate analysis.

Thank you for your answers. A clustering with PAST or R seems appropriate. However, which clustering index is the most appropriate knowing that my binary matrix corresponds at the presence/absence of genes in a set of genomes. I know it was discussed here: Similarity Measures Appropriate For Hierarchical Clustering On Gene Content/Binary Data

but I want to know your opinions.