Polyphyletic groups are overwriting coloring in phylogenetic trees

[croots]

I believe I've encountered a bug, or at least undocumented behavor. The phylogenetic tree seems to enforce that groups are at least paraphyletic. If I have polyphyletic coloring groups, they begin to overwrite each other, producing unexpected color outputs.

Minimum reproducible code forthcoming

[croots]

from re import PatternError
from pycirclize import Circos
from pycirclize.utils import load_example_tree_file, ColorCycler
from matplotlib.lines import Line2D
from Bio import Phylo
from Bio import Phylo
from io import StringIO

newick_str = "((p1,(p2,p3,p4),(p5,p6,p7)),p8);"

handle = StringIO(newick_str)
tree = Phylo.read(handle, "newick")

circos, tv = Circos.initialize_from_tree(
    tree,
    start=-90,
    end=270,
    r_lim=(30, 100),
    leaf_label_size=5,
    ignore_branch_length=True,
    label_formatter=lambda t: t.replace("_", " "),
)

# Define group-species dict for tree annotation
# In this example, set minimum species list to specify group's MRCA node
colors = {
    "group 1": ["p1", "p3", "p4"],
    "group 2": ["p2", "p5", "p6"],
    "group 3": ["p7", "p8"]
}

# Set tree line color & label color
ColorCycler.set_cmap("tab10")
group_name2color = {name: ColorCycler() for name in colors.keys()}
for group_name, species_list in colors.items():
    color = group_name2color[group_name]
    tv.set_node_line_props(species_list, color=color, apply_label_color=True)

# Plot figure & set legend on center
fig = circos.plotfig()
_ = circos.ax.legend(
    handles=[Line2D([], [], label=n, color=c) for n, c in group_name2color.items()],
    labelcolor=group_name2color.values(),
    fontsize=6,
    loc="center",
    bbox_to_anchor=(0.5, 0.5),
)

fig.savefig("bug.png")

[moshi4]

The code above is working as intended and is not a bug.

The tv.set_node_line_props method sets the color for the MRCA(Most Recent Common Ancestor) node of the group species list and all its descendant nodes. Therefore, overwriting the color for the root node(= MRCA node for group3=[“p7”, “p8”] in the example) and all its descendant nodes is the correct behavior.

I don't fully understand what your ideal figure looks like, but achieving polyphyletic coloring groups with pyCirclize might be a bit difficult.

[croots]

I think my expectations would have been that each leaf is colored in the group assigned to them explicitly by the code provided. For example, only p7 and p8 would display green here, with ["p1", "p3", "p4"] and ["p2", "p5", "p6"] being blue and orange, respectively. From there, it would probably look like the documentation's example, with nodes that have multiple groups being black. Alternatively, it could solve for a most parsimonious solution and have ancestral nodes colored accordingly.

If this is 'will not change', a note in the documentation that polyphyletic groups aren't supported would be beneficial so people that need them know immediately to look elsewhere.

In either case, thanks for supporting this tool. I'm pivoting to ete4 for my phylogenetics since I need something more robust, but your circos and comparative genomics tools look super useful as well