Columns order in Oncoprint

[jfrigola]

Hi everyone! I'm using the oncoPrintPlotter and I'd like to be able to sort columns by myself. Looking at the previous discussion, I've seen you suggest using "col_split" and then "col_split_order" to specify the order I want. However when I do this I'm not able to remove the grid and spines of the top annotations as you describe in the documentation (it does work but only on the last column, all other column annotations keep their grid and spines):

#remove the grid
op.top_annotation.annotations[0].ax.grid(False)
#remove spines for top annotation and right annotation
despine(ax=op.top_annotation.annotations[0].ax,left=False, bottom=True, right=False, top=True)
despine(ax=op.right_annotation.annotations[0].ax,left=True, bottom=False, right=True, top=False)

Is there a better way to order samples in an oncoprint or a way I could remove grid and spines if the only way to do it is by using col_split and col_split order?

Also, columns can be easily sorted when using ClusterMapPlotter, but if I want my heatmap to display categorical data, the oncoprint should be used, right?

[DingWB]

Thanks for your question @jfrigola . If you want to remove the grid and spine for all annotations, then you need to write a for loop.
In the documentation, I only remove the grid and spine for the first annotation (Because I had only one annotation in the documentation, op.top_annotation.annotations[0], 0 means the first annotation).
You can remove the grid and spine for all annotation axes by writing a for loop:

for annotation in op.top_annotation.annotations:
    ax=annotation.ax
    ax.grid(False)
    #remove spines for top annotation and right annotation
    despine(ax=ax,left=False, bottom=True, right=False, top=True)
    despine(ax=ax,left=True, bottom=False, right=True, top=False)

I hope this will solve your issue.

[jfrigola]

Thanks for your reply! This code works fine if I don't use col_split. However, if I use col_split, then the for loop iterates all annotations as it should, but only removes the grid and spine of the last column split of each annotation.

[DingWB]

Maybe you can try to set each axes in top_annotation.axes, which contains all axes if row_split or col_split was used.
You can also paste your code and dataset and we can debug it for you.

[jfrigola]

Sure, I've prepared a toy example so that you can see what I refer to:

(By the way, I'm facing this problem because using col_split is the only way I've found to be able to sort columns in an oncoprint. As a suggestion, I believe it would be nice to be able to sort columns in an easier way. In any case though, this package is very nice and useful, thanks for your work!)

import pandas
import random

# Create toy dataset
samples = [f'Sample_{i}' for i in range(1, 11)]
genes = [f'Gene_{i}' for i in range(1, 11)]

alts_lol = list()

for sample in samples:
    for gene in genes:
        amp_value = random.randint(0, 1)
        
        if amp_value == 0:
            del_value = random.randint(0, 1)
            
        else:
            del_value = 0
            
        if (amp_value == 0) & (del_value == 0):
            neut_value = 1
            
        else:
            neut_value = 0
            
        alts_lol.append([sample, gene, amp_value, neut_value, del_value])

alts_df = pd.DataFrame(alts_lol, columns=['sample', 'gene', 'amp', 'neut', 'del'])

# Prepare column annotations
annot_1_df = pd.DataFrame([[i, random.randint(0,100)] for i in samples], columns=['sample', 'annot1'])
annot_1_df.index = annot_1_df['sample']

annot_2_df = pd.DataFrame([[i, random.randint(500, 5000)] for i in samples], columns=['sample', 'annot2'])
annot_2_df.index = annot_2_df['sample']

annot_3_df = pd.DataFrame([[i, 'patient' + str(random.randint(1,5))] for i in samples], 
                            columns=['sample', 'patient'])
annot_3_df.index = annot_3_df['sample']

top_annotation=HeatmapAnnotation(A3=anno_simple(annot_3_df['patient']),
                                 A1=anno_barplot(annot_1_df['annot1']),
                                 A2=anno_barplot(annot_2_df['annot2']))

# Plot oncoprint
plt.figure(figsize=(3,4))

op=oncoPrintPlotter(data=alts_df, y='gene', x='sample', values=['amp', 'neut', 'del'], 
                    show_rownames=True, show_colnames=True, colors=['red', 'white', 'green'],
                    top_annotation=top_annotation, col_split=annot_3_df['patient'], col_split_gap=1)

# Remove the grid
for annotation in op.top_annotation.annotations:
    ax=annotation.ax
    ax.grid(False)
    #remove spines for top annotation and right annotation
    despine(ax=ax,left=False, bottom=True, right=False, top=True)
    despine(ax=ax,left=True, bottom=False, right=True, top=False)

[DingWB]

Is this what you want?

from PyComplexHeatmap import *
import pandas as pd
import random

# Create toy dataset
samples = [f'Sample_{i}' for i in range(1, 11)]
genes = [f'Gene_{i}' for i in range(1, 11)]

alts_lol = list()

for sample in samples:
    for gene in genes:
        amp_value = random.randint(0, 1)
        
        if amp_value == 0:
            del_value = random.randint(0, 1)
            
        else:
            del_value = 0
            
        if (amp_value == 0) & (del_value == 0):
            neut_value = 1
            
        else:
            neut_value = 0
            
        alts_lol.append([sample, gene, amp_value, neut_value, del_value])

alts_df = pd.DataFrame(alts_lol, columns=['sample', 'gene', 'amp', 'neut', 'del'])

# Prepare column annotations
annot_1_df = pd.DataFrame([[i, random.randint(0,100)] for i in samples], columns=['sample', 'annot1'])
annot_1_df.index = annot_1_df['sample']

annot_2_df = pd.DataFrame([[i, random.randint(500, 5000)] for i in samples], columns=['sample', 'annot2'])
annot_2_df.index = annot_2_df['sample']

annot_3_df = pd.DataFrame([[i, 'patient' + str(random.randint(1,5))] for i in samples], 
                            columns=['sample', 'patient'])
annot_3_df.index = annot_3_df['sample']


top_annotation=HeatmapAnnotation(A3=anno_simple(annot_3_df['patient'],
                                               add_text=True,height=5,
                                                text_kws={'color':'white', 'fontsize':4,'fontweight':'bold',
                                                          'rotation':0,'horizontalalignment':'left'}),
                                 A1=anno_barplot(annot_1_df['annot1']),
                                 A2=anno_barplot(annot_2_df['annot2']))

# Plot oncoprint
plt.figure(figsize=(4,6))

op=oncoPrintPlotter(data=alts_df, y='gene', x='sample', values=['amp', 'neut', 'del'], 
                    show_rownames=True, show_colnames=True, colors=['red', 'white', 'green'],
                    top_annotation=top_annotation, col_split=annot_3_df['patient'], 
                    col_split_gap=0.2,width=0.8) #width control the width of the bar in each cell

# Remove the grid
# for annotation in op.top_annotation.annotations:
#     ax=annotation.ax
#     ax.grid(False)
#     #remove spines for top annotation and right annotation
#     despine(ax=ax,left=False, bottom=True, right=False, top=True)
#     despine(ax=ax,left=True, bottom=False, right=True, top=False)

for ax in op.top_annotation.axes.flatten():
    ax.grid(False)
    #remove spines for top annotation and right annotation
    despine(ax=ax,left=False, bottom=True, right=False, top=True)
    despine(ax=ax,left=True, bottom=False, right=True, top=False)

[DingWB]

I have already added this example onto the documentation website: https://dingwb.github.io/PyComplexHeatmap/build/html/notebooks/oncoPrint2.html
I hope you don't mind.

[jfrigola]

That's what I trying to do. Thanks! :)

[DingWB]

Thanks for letting me know @jfrigola