Replicate genome dendrogram about anvio

[meren]

Hey @lucaz88,

The clustering of genomes in the pangenomic workflow takes place here in panops.py:

(...)
for clustering_tuple in [('gene_cluster presence absence', self.view_data),
                         ('gene_cluster frequencies', self.view_data_presence_absence)]:
    v, d = clustering_tuple
    newick = clustering.get_newick_tree_data_for_dict(d, 
                                                      transpose=True,
                                                      distance = self.distance,
                                                      linkage=self.linkage)
(...)

If you follow the function get_newick_tree_data_for_dict in panops.py you will find that the default clustering uses 'euclidean' distance with 'ward' linkage on l1-normalized vectors by default, where linkage is calculated like this:

linkage = hierarchy.linkage(vectors, metric=distance, method=linkage)

Your first dendrogram is almost identical to the one in anvi'o (if you rotate things around a bit), and differences are likely due to the normalization step. I hope this helps.

Best wishes,
Meren.

[lucaz88]

Thanks @meren !! I missed the normalization step! Now it looks identical without the need to do rotate anything 😉

[meren]

Perfect :) Thank you for reporting back!