Inserting gaps between rows and/or columns with matshow

Question

I am hoping to introduce gaps between specific rows / columns in heatmap similar to R's pheatmap gaps_row or gaps_col argument. However, after a cursory search this feature does not appear to be available. I have considered emulating it by creating different axes with specific sizes to emulate this behavior but getting their placement correct would be tricky. Is there an easier way I could go about emulating this functionality?

For example:

From this question

ImportanceOfBeingErnest · Accepted Answer · 2017-04-13T16:36:20.823

Here is a code to reproduce the above figure in matplotlib.

import matplotlib.pyplot as plt
import numpy as np; np.random.seed(0)

a = np.random.poisson(lam=5, size=(10*5, 4*3))

fig, axes = plt.subplots(nrows=5, ncols=4+1, figsize=(6.5,7),
                         gridspec_kw={"width_ratios":4*[1] + [0.2]})

kw = dict(aspect="auto",vmin=a.min(), vmax= a.max())
for i in range(5):
    for j in range(4):
        im = axes[i,j].imshow(a[10*i:10*i+10,3*j:3*j+3], **kw )
        axes[i,j].tick_params(axis=u'both', which=u'both',length=0)
        axes[i,j].set_xticklabels([])
        axes[i,j].set_yticklabels([])

ylabels=["gene {:02d}".format(i+1) for i in range(50)]
xlabels=["treatment {:02d}".format(i+1) for i in range(12)]
clabels=["{:02d}".format(i+1) for i in range(5)]
for i in range(5):
    axes[i,4].set_facecolor(plt.cm.Set2(i/8.))
    axes[i,4].text(0.5,.5, clabels[i], rotation=-90, color="w",
                   transform=axes[i,4].transAxes,
                   ha="center", va="center", fontweight="bold", fontsize=9)
    axes[i,4].tick_params(axis=u'both', which=u'both',length=0)
    axes[i,4].set_xticklabels([])
    axes[i,4].set_yticklabels([])
    axes[i,0].set_yticks(range(10))
    axes[i,0].set_yticklabels(ylabels[i*10:i*10+10], fontsize=7)

for j in range(4):
    axes[4,j].set_xticks(range(3))
    axes[4,j].set_xticklabels(xlabels[j*3:j*3+3], fontsize=9, rotation=90)
axes[4,4].set_xticks([0.5])
axes[4,4].set_xticklabels(["category"], fontsize=9, rotation=90, fontweight="bold")    

cax = fig.add_axes([0.9,0.5,0.03,0.44])
fig.colorbar(im, cax=cax) 


plt.subplots_adjust(bottom=0.2, top=0.94, right=0.86)   
plt.show()

If not all of the rows or columns have the same number of datapoints, but if data points are the same along one row or column, this solution would still work. You would then need to adjust the height_ratios or width_ratios of the gridspec, e.g. if the second column has 5 datapoints along x direction instead of 3, you would call

gridspec_kw={"width_ratios":[3,5,3,3,1]}

Ofg course the loops would need to be adapted to account for the different number of ticks etc.

This is great thanks. If the columns or rows are of unequal sizes would this still work while preserving the size of each data point? — GWW, Apr 13 '17 at 16:02
If I understand correctly you want to have e.g. 15 datapoints instead of 10 in one of the rows? That would work, as long as the complete row has the same number of points. Same for the columns. I updated the answer with a paragraph about this. — ImportanceOfBeingErnest, Apr 13 '17 at 16:36
That's really helpful thanks. I didn't know about the width and height ratios for the gridspec. I have been drawing plots like this by manually adding each axes using `add_axes` — GWW, Apr 14 '17 at 19:55

Inserting gaps between rows and/or columns with matshow

1 Answers1