Custom pcolor in matplotlib

Question

I am plotting several heatmaps in matplotlib as shown below.

Here is my loop:

with open(gene_peak) as f:
    count = 1
    for line in f:
        np_array=[]
        gene_peak = line.strip().split("\t")
        gene_id = gene_peak[0]
        peaks = gene_peak[1].split(",")
        for peak in peaks:
            np_array.append(enhancer_fc[peak])
        data, pval = stats.spearmanr(np.transpose(np.array(np_array)))
        plt.subplot(4,3,count+1)
#       plt.title(gene_id)
        plt.pcolor(data, cmap=plt.cm.OrRd, vmin=-1, vmax=1)
        plt.gca().invert_yaxis()
        plt.gca().set_aspect(aspect='equal', adjustable='box-forced')
        plt.xticks([])
        plt.yticks([])
        print count
        count += 1
    plt.show()

I am plotting the spearman correlations of different 2D arrays of different dimensions.

Question:

There are correlation values, so they range from -1 to 1. I want to add custom colorbar() such that values above 0.4 starts showing a gradient of red and below -0.4 shows a gradient of blue, such that I show only the points that are more than 0.4 and less than -0.4.

Also I would like to plot only one colorbar() such that the image looks cleaner. Any help would be appreciated, Thanks.

Answer 1

You can define your own discrete colormap using the ListedColorMap from Matplotlib. You can use the colorbar from one of the plots, and place it in position so it represent all of the plots visually. Here is an example with the colours you have given:

from matplotlib import colors


discrete_colors = [(255, 0, 20), (255, 70, 65), (255, 128, 110), (255, 181, 165), (64, 64, 64),
    (0, 0, 0), (64, 64, 64), (124, 128, 217), (102, 107, 216), (69, 76, 215), (33, 33, 245)]
discrete_colors = [(r/255., g/255., b/255.) for r, g, b in discrete_colors]         

my_colormap = colors.ListedColormap(discrete_colors)

subplot(211)
data = 2 * np.random.rand(10, 10) - 1.0
pcolor(data, cmap=my_colormap, vmin=-1, vmax=1)
subplot(212)  # Some other plot
data = 2 * np.random.rand(10, 10) - 1.0
pc = pcolor(data, cmap=my_colormap, vmin=-1, vmax=1)

fig = gcf()
fig.subplots_adjust(right=0.70)
cax = fig.add_axes([0.80, 0.15, 0.05, 0.7])
fig.colorbar(pc, cax=cax)

You might have to adjust the code a bit. I'm using IPython 2.7.