additional row colors in seaborn cluster map

Question

I am currently generating clustermaps in seaborn and labeling the row colors as below.

matrix = pd.DataFrame(np.random.random_integers(0,1, size=(50,4)))
labels          


        

Answer 1

The solution is below. The seaborn API does actually allow this to be done.

matrix = pd.DataFrame(np.random.random_integers(0,1, size=(50,4)))

labels = np.random.random_integers(0,5, size=50)
lut = dict(zip(set(labels), sns.hls_palette(len(set(labels)), l=0.5, s=0.8)))
row_colors = pd.DataFrame(labels)[0].map(lut)

#Create additional row_colors here
labels2 = np.random.random_integers(0,1, size=50)
lut2 = dict(zip(set(labels2), sns.hls_palette(len(set(labels2)), l=0.5, s=0.8)))
row_colors2 = pd.DataFrame(labels2)[0].map(lut2)

g=sns.clustermap(matrix, col_cluster=False, linewidths=0.1, cmap='coolwarm', row_colors=[row_colors, row_colors2])
plt.show()

This produces a Clustermap with two additional columns:

Answer 2

I tried to concat the row_colors dataframe by pandas and it worked! Please try this code:

import seaborn as sns; sns.set(color_codes=True)
import matplotlib.pyplot as plt
import pandas as pd

iris = sns.load_dataset("iris")
print(iris)
species = iris.pop("species")


lut1 = dict(zip(species.unique(), ['#ED2323','#60FD00','#808080']))
row_colors1 = species.map(lut1)

lut2 = dict(zip(species.unique(), "rbg"))
row_colors2 = species.map(lut2)

row_colors = pd.concat([row_colors1,row_colors2],axis=1)
print(row_colors)

g = sns.clustermap(iris, row_colors=row_colors, col_cluster=False,cmap="mako", yticklabels=False, xticklabels=False)

plt.show()

Answer 3

There is another option for feeding in the annotation colors: you can provide a whole dataframe in the row colors or col_colors options, instead of a list of lists.

This strategy might be particularly helpful if you have a dataframe with several annotations you want represented. Instead of map, you can use the pandas function replace.

Something such as this bit can be used to modify the other answer:

## This step is necessary because you can't use replace with the tuple rgb values
lut = {k:matplotlib.colors.to_hex(v) for k, v in lut.iteritems()}

annotations_df = annotations_df.replace(lut)

g=sns.clustermap(matrix, col_cluster=False, linewidths=0.1, cmap='coolwarm', row_colors=annotations_df)
plt.show()