clustermap: tight_layout hardocoded parameters

[dmaziec]

Hello!

Thank you for your amazing package that makes data science easier!

I am using the clustermap method and run into problems related the color bar that overlapped the heat map. In order to increase the distance between the heat map and the color bar I would like to increase the padding that is set in matplotlib.pyplot.tight_layout, the h_pad and w_pad parameters. Currently, these values are hardcoded and I believe that makes the color bar overlap one of the dendrograms sometimes:

seaborn/seaborn/matrix.py

Line 1105 in 4f72554

tight_params = dict(h_pad=.02, w_pad=.02)

Please see my attached picture below. When I increased the padding, it helped me and there was no overlap.

Would it be possible to have the parameters of tight_layout accessible through clustermap ?
Otherwise, what would be the other recommended solution to solve this problem?

Thank you,
Dominika

[mwaskom]

Can you get the result you're looking for by tweaking cbar_pos? (It would be easier to help with a reproducible example).

[dmaziec]

I am running my analysis for 40 different datasets so I would prefer a more generic, automated way instead of using cbar_pos.

Here is my code example:
Please download this dataset and run the Python script posted below from the same folder where the file is placed: Stomach-AdenoCA.csv

The output heat map should be saved in Stomach-AdenoCA.pdf.

import pandas
import matplotlib.pyplot as plt
import seaborn as sns

cluster_data = pandas.read_csv(
        "Stomach-AdenoCA.csv",
        index_col=False,
    )
his="Stomach-AdenoCA"
uuids = cluster_data["UUID"]
cluster_map = sns.clustermap(
                cluster_data.drop("UUID", axis=1),
                cmap=sns.color_palette("crest", as_cmap=True),  # green colors
                yticklabels=list(uuids),  # uuids labels instead of indexes
            )


cluster_map.ax_heatmap.set_yticklabels(
    cluster_map.ax_heatmap.get_yticklabels(),
    fontdict={"fontsize": 8},
)
cluster_map.ax_heatmap.set_xticklabels(
    cluster_map.ax_heatmap.get_xticklabels(), fontdict={"fontsize": 8}
)
cluster_map.ax_heatmap.set_ylabel("Sample UUID")
cluster_map.ax_heatmap.set_xlabel("Rearrangement type")
cluster_map.ax_heatmap.tick_params(axis="y", width=0.5)
cluster_map.figure.suptitle(
    f"Clustering of structural variant patterns of {his} samples.",
    size="xx-large",
    x=0.55,
    y=1,
)
cluster_map.ax_col_dendrogram.set_visible(False)

cluster_map.savefig(f"{his}.pdf")

plt.close()

To solve my problem I changed the parameters of tight_layout following the implementation in the clustermap function . This is the code I reused but I changed the value of h_pad. I added this piece of code before saving the map to the file (savefig) :

# to avoid cbar overlapping with the dendrogram
cluster_map.ax_cbar.set_axis_off()
cluster_map.figure.tight_layout(h_pad=1, w_pad=0.02, rect=[0, 0, 1, 1])
cluster_map.ax_cbar.set_axis_on()
cluster_map.ax_cbar.set_position((0.02, 0.8, 0.05, 0.18))

I don't struggle with this anymore since I figured how to do it on my end so feel free to close the issue.