[FIXED] Matplotlib for multi-panel of subgraphs

Issue

I would like to plot multi-panel graphs using the following data:

CHROM   BIN_START   BIN_END N_VARIANTS  PI

NODE_10_length_497_cov_170.299805   1   10000   39  0.000703581
NODE_13_length_1438_cov_357.659943  1   10000   61  0.00132786
NODE_15_length_733_cov_44.686222    1   10000   25  5.73165e-05
NODE_16_length_8151_cov_58.001228   1   10000   525 0.0129994
NODE_18_length_98839_cov_51.306377  1   10000   753 0.0158054
NODE_18_length_98839_cov_51.306377  10001   20000   644 0.0147527
NODE_18_length_98839_cov_51.306377  20001   30000   783 0.0180735
NODE_18_length_98839_cov_51.306377  30001   40000   433 0.00950069
NODE_18_length_98839_cov_51.306377  40001   50000   568 0.0134851
NODE_18_length_98839_cov_51.306377  50001   60000   634 0.0107662
NODE_18_length_98839_cov_51.306377  60001   70000   501 0.0104874
NODE_18_length_98839_cov_51.306377  70001   80000   736 0.0142129
NODE_18_length_98839_cov_51.306377  80001   90000   803 0.0177756
NODE_18_length_98839_cov_51.306377  90001   100000  683 0.015499
NODE_20_length_219211_cov_54.980137 1   10000   803 0.0188128
NODE_20_length_219211_cov_54.980137 10001   20000   577 0.0152466
NODE_20_length_219211_cov_54.980137 20001   30000   596 0.0149394
NODE_20_length_219211_cov_54.980137 30001   40000   982 0.0230532
NODE_20_length_219211_cov_54.980137 40001   50000   991 0.0211885
NODE_20_length_219211_cov_54.980137 50001   60000   659 0.0154533
NODE_20_length_219211_cov_54.980137 60001   70000   752 0.0171805
NODE_20_length_219211_cov_54.980137 70001   80000   725 0.0168676
NODE_20_length_219211_cov_54.980137 80001   90000   836 0.0199372
NODE_20_length_219211_cov_54.980137 90001   100000  652 0.0116497
NODE_20_length_219211_cov_54.980137 100001  110000  788 0.0176188
NODE_20_length_219211_cov_54.980137 110001  120000  278 0.00549361
NODE_20_length_219211_cov_54.980137 120001  130000  499 0.0120812
NODE_20_length_219211_cov_54.980137 130001  140000  738 0.0161006
NODE_20_length_219211_cov_54.980137 140001  150000  729 0.0148927
NODE_20_length_219211_cov_54.980137 150001  160000  611 0.015047
NODE_20_length_219211_cov_54.980137 160001  170000  737 0.0175226
NODE_20_length_219211_cov_54.980137 170001  180000  709 0.0153575
NODE_20_length_219211_cov_54.980137 180001  190000  688 0.0133387
NODE_20_length_219211_cov_54.980137 190001  200000  693 0.0145321
NODE_20_length_219211_cov_54.980137 200001  210000  564 0.0122298
NODE_20_length_219211_cov_54.980137 210001  220000  687 0.0140385
NODE_22_length_248_cov_293.927429   1   10000   23  0.000482139
NODE_24_length_104_cov_29.778847    1   10000   6   2.73246e-05
NODE_29_length_319744_cov_53.127953 1   10000   759 0.0155234
NODE_29_length_319744_cov_53.127953 10001   20000   741 0.0182896
NODE_29_length_319744_cov_53.127953 20001   30000   709 0.0147025
NODE_29_length_319744_cov_53.127953 30001   40000   755 0.0159811
.
.
.

I have used the following python code.

    import matplotlib.pyplot as plt

    chrs = {}

    with open('diversity_level.windowed.pi', 'r') as f:
        info = f.readlines()

    for i in range(2, len(info)):#skip the first line
        linfo = info[i].split("\t")
        if (linfo[0] in chrs):
        chrs[linfo[0]][0].append(int(linfo[1]))
        chrs[linfo[0]][1].append(float(linfo[4]))
    else:
        chrs[linfo[0]] = [[] for i in range(2)]

    fig, axs = plt.subplots(len(chrs), 1)
    fig.subplots_adjust(wspace=50,hspace=0.1)

    i = 0
    for chr, data in chrs.items():
        axs[i].plot(data[0], data[1])
        axs[i].set_xlabel(chr)
        axs[i].set_ylabel('pi')
        axs[i].grid(True)
        i = i + 1

    plt.show()

The obtained result is

May I know how to adjust the parameter in the code so the the X-axis label of each subgraph can be seen. How can I select not to output empty subgraphs. Thanks in advance

To gobffi, Since all the data is inside the file 'diversity_level.windowed.pi', how can I read the data in the file (without header) and put in the dictionary for plotting the subgraph?

from matplotlib.pyplot import show, subplots
data_dict = {}

with open('diversity_level.windowed.pi', 'r') as f:
    data= f.readlines()
   

for record in range(1,len(data)): #skip the first line header
    key, start, pi = record[0], int(record[1]), float(record[4])
    start_list, pi_list = data_dict.get(key, [[],[]])
    start_list.append(start), pi_list.append(pi)
    data_dict[key] = [start_list, pi_list]
for key in list(data_dict.keys()):
    if len(data_dict[key][0])==1: data_dict.pop(key)

fig, axs = subplots(len(data_dict), constrained_layout=1,
                    sharex=True, sharey=True)

for ax, (key, (start_list, pi_list)) in zip(axs, (data_dict.items())):
    ax.plot(start_list, pi_list)
    ax.set_xlabel(key, size='x-small')
    ax.set_ylabel('pi', size='x-small')
    ax.tick_params(axis='both', labelsize='x-small')
    ax.grid(1);
ax.set_ylim(bottom=0.0)
show()

Solution

Here it is my attempt, note that I have reorganized your code a little bit because I was feeling overwhelmed by the amount of indices you used.

1. To remove the "empty" sequences I loop over the keys, and if the length of a sequence is exactly 1 I remove (that is, pop) the corresponding dictionary entry.
2. To solve the issue with the cramped plots, usually a good idea is to specify that the figure has to respect constrained_layout (it's a relatively new feature that, imho, was really missing!)

.

from matplotlib.pyplot import show, subplots

# iterate over the lines of a file opened for reading
# each line is a record
data = [record.split() for record in open('your_file_name', 'r')]
    
data_dict = {}

# we iterate on all the records, except the first one (Python counts from 0)
for record in data[1:]:

    # unpack the "interesting" stuff
    key, start, pi = record[0], int(record[1]), float(record[4])

    # get what is inside the dictionary item, using unpacking
    # if item is "new" we get the default value, that is two empty lists
    start_list, pi_list = data_dict.get(key, [[],[]])
    
    # append the values to the respective lists
    start_list.append(start), pi_list.append(pi)

    # update the dictionary item with the augmented lists
    data_dict[key] = [start_list, pi_list]

# we remove the "non interesting" entries from the dictionary
# using the `.pop()` method
# note the use of `list(...)` to have a static reference to the keys
for key in list(data_dict.keys()):
    if len(data_dict[key][0])==1: data_dict.pop(key)

# instantiate the figure and the axes, using `constrained_layout`
# usually leads to a better arrangement of the figure elements
fig, axs = subplots(len(data_dict), constrained_layout=1)

# except for a moderate abuse of the unpacking syntax,
# everything should be clear …   
for ax, (key, (start_list, pi_list)) in zip(axs, (data_dict.items())):
    ax.plot(start_list, pi_list)
    ax.set_xlabel(key)
    ax.set_ylabel('pi')
    ax.grid(1);
show()

---

A twist on the subject, if the data instances must be compared I prefer to have the same axes (hence sharex and shareyinsubplots`) in each subplot and also, in this particular case, to set the lower y limit to zero; also, because the figure is rather busy, here I show you how to reduce the size of the labels and of the tick labels

...

fig, axs = subplots(len(data_dict), constrained_layout=1,
                    sharex=True, sharey=True)

for ax, (key, (start_list, pi_list)) in zip(axs, (data_dict.items())):
    ax.plot(start_list, pi_list)
    ax.set_xlabel(key, size='x-small')
    ax.set_ylabel('pi', size='x-small')
    ax.tick_params(axis='both', labelsize='x-small')
    ax.grid(1);
ax.set_ylim(bottom=0.0)
show()

Answered By - gboffi