In [8]:
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import matplotlib.patches as mpatches
# Sort the DataFrame by 'Condition' and 'line'
Compositions_sorted = Compositions.sort_values(["Condition", "line"])
# Ensure the sample order is preserved as a categorical type
Compositions_sorted["sample_id"] = pd.Categorical(
Compositions_sorted["sample_id"],
Compositions_sorted["sample_id"].unique().tolist()
)
# Plot with sorted x-axis
fig, ax = plt.subplots(1, 1, figsize=(30, 7), dpi=100)
sns.histplot(
data=Compositions_sorted,
x="sample_id",
hue="phase",
weights="SampleNormSize",
multiple="stack",
shrink=0.8,
ax=ax
)
ax.set_ylabel("Cells Fraction")
ax.set(ylim=(0, 1))
sns.despine(ax=ax)
# Try to retrieve the auto-generated legend handles/labels for phase
macro_handles, macro_labels = ax.get_legend_handles_labels()
# Remove the auto legend to avoid conflicts.
if ax.get_legend() is not None:
ax.get_legend().remove()
# If macro_handles is empty, build the legend manually.
if len(macro_handles) == 0:
macro_unique = Compositions_sorted["phase"].unique()
# Use a palette (here we use "deep" which is commonly used by seaborn)
macro_palette = sns.color_palette("deep", n_colors=len(macro_unique))
macro_dict = {macro: color for macro, color in zip(macro_unique, macro_palette)}
macro_handles = [mpatches.Patch(color=color, label=macro) for macro, color in macro_dict.items()]
macro_labels = list(macro_dict.keys())
# Group info by sample_id for Condition and line
group_info = Compositions_sorted.groupby("sample_id").agg({"Condition": "first", "line": "first"})
# Get x positions for categorical data
xticks = ax.get_xticks()
samples_order = list(Compositions_sorted["sample_id"].cat.categories)
sample_to_x = {sample: xtick for sample, xtick in zip(samples_order, xticks)}
# Assign colors for Conditions and lines
unique_conditions = group_info["Condition"].unique()
unique_lines = group_info["line"].unique()
condition_colors = {
cond: color for cond, color in zip(unique_conditions, sns.color_palette("tab10", n_colors=len(unique_conditions)))
}
line_colors = {
line: color for line, color in zip(unique_lines, sns.color_palette("pastel", n_colors=len(unique_lines)))
}
# Use a transform that uses data coordinates for x and axes coordinates for y
trans = ax.get_xaxis_transform()
# Add colored bars above the main plot for each sample
for sample, row in group_info.iterrows():
x_center = sample_to_x[sample]
rect_width = 1 # each category is assumed to be width 1
ax.add_patch(mpatches.Rectangle((x_center - 0.5, 1.02), rect_width, 0.02,
color=condition_colors[row["Condition"]],
transform=trans,
clip_on=False))
ax.add_patch(mpatches.Rectangle((x_center - 0.5, 1.05), rect_width, 0.02,
color=line_colors[row["line"]],
transform=trans,
clip_on=False))
# Create custom legends for Condition and line groupings
condition_legend = [mpatches.Patch(color=color, label=cond) for cond, color in condition_colors.items()]
line_legend = [mpatches.Patch(color=color, label=line) for line, color in line_colors.items()]
# Adjust the right margin to make room for legends
fig.subplots_adjust(right=1.5)
# Place legends outside the main axes
legend_macro = ax.legend(handles=macro_handles, labels=macro_labels, title="Phase", loc="upper left", bbox_to_anchor=(0, 1))
ax.add_artist(legend_macro)
legend_condition = ax.legend(handles=condition_legend, title="Condition", loc="upper left", bbox_to_anchor=(.955, .9))
ax.add_artist(legend_condition)
legend_line = ax.legend(handles=line_legend, title="Line", loc="upper left", bbox_to_anchor=(.955, 1.1))
ax.add_artist(legend_line)
ax.grid(False)
plt.xticks(rotation=90)
plt.savefig(f"./figures/{DS}.Prefilt.scaled.CyclingRate.svg", bbox_inches="tight")
plt.show()
