Add ylim to anno_scatterplot

[Laolga]

Hi! Could you please advice how to add ylim to anno_scatterplot?

[DingWB]

As for now, you can get the annotation for anno_scatterplot and set the ylim, for example:

ann=col_ha.annotations[0]:
ann.ax.set_ylim([0,1])

[Laolga]

Thanks!

[zcfu21]

Hi! A similar question in 'anno_barplot': how to set 'yticks' to anno_barplot? I've tried 'row_ha.annotations[0].ax.set_yticks()', but it raises an error ('anno_barplot' object has no attribute 'ax').

[DingWB]

What is your row_ha, row_ha.annotations[0] means the first element of row_ha.annotations, if the barplot is not the first element, please change 0 to the corresponding int number.
Why don't you paste all of your code so I can look at and debug it for you?

[DingWB]

According to the source code:

  if axis == 1:
      ax.set_xticks(ticks=np.arange(0.5, self.nrows, 1))
      ax.bar(
          x=np.arange(0.5, self.nrows, 1),
          height=self.plot_data[col].values,
          bottom=base_coordinates,
          color=color,
          **plot_kws
      )
  else:
      ax.set_yticks(ticks=np.arange(0.5, self.nrows, 1))
      ax.barh(
          y=np.arange(0.5, self.nrows, 1),
          width=self.plot_data[col].values,
          left=base_coordinates,
          color=color,
          **plot_kws
      )

You can set custom yticks for column annotations and xticks for row annotations.
Why do you want to set_yticks for row annotations? yticks for row annotations should be shared by all annotations and should be a range from 0 to nrows, no need to set yticks for row annotation. @zichufu

Maybe you wanted to set_xticks for row annotations?

[zcfu21]

Sorry for my unclear descriptions. I want to change the yticks/grids in the column annotation barplot from [0,2.5] to [0,1,2,3] (see the following figure). Is there any way to do this? Thanks you so much.

[DingWB]

I would suggest you try:
(1). set_ylim and set_yticks
(2).
ax=col_ha.annotations[0].ax
ax.yaxis.set_major_locator(MultipleLocator(1))

Please let me know whether it works or not.

[zcfu21]

Here are my codes:

with plt.rc_context({'font.sans-serif': ['Arial']}):
    plt.figure(figsize=(15, 15))
    col_ha = HeatmapAnnotation(GeneNums=anno_barplot(df_bar,
                                           height=10,colors='orangered'),legend=True,legend_gap=5,hgap=0.5)
    ax=col_ha.annotations[0].ax
    ax.yaxis.set_major_locator(MultipleLocator(1))
    row_ha = HeatmapAnnotation(
                            selected=anno_label(label_rows,colors=colors_dict,relpos=(0,0.4)),
                            axis=0,verbose=0,orientation='right'
                            )
    cm = ClusterMapPlotter(data=distance_df,top_annotation=col_ha,right_annotation=row_ha,
                           col_split=6, row_split=6, col_split_gap=1,
                           row_cluster_metric='euclidean',
                           col_cluster_metric='euclidean',
                           row_cluster_method='ward',
                           col_cluster_method='ward',
                           row_split_gap=1,label='distance',
                           row_dendrogram=True,col_dendrogram=False,show_rownames=False,show_colnames=True,
                           tree_kws={'row_cmap': 'Dark2'},cmap='Spectral_r',
                           legend_gap=5,legend_hpad=2,legend_vpad=5)
    plt.show()

The error is

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
~\AppData\Local\Temp\ipykernel_20932\4202014503.py in <module>
      4                                            height=10,colors='orangered'),legend=True,legend_gap=5,hgap=0.5)
      5     #col_ha.ax.set_yticks([0,1,2,3])
----> 6     ax=col_ha.annotations[0].ax
      7     ax.yaxis.set_major_locator(MultipleLocator(1))
      8 

AttributeError: 'anno_barplot' object has no attribute 'ax'

[DingWB]

please try to put

ax=col_ha.annotations[0].ax
ax.yaxis.set_major_locator(MultipleLocator(1))

after ClusterMapPlotter.
Because the axes will be created only after you plot the heatmap.

[zcfu21]

Ohhh it works. Thank you so much!

[DingWB]

That's great. Your heatmap looks beautiful.
Could you please paste your code and heatmap here so that other users will know how to make it work if they have the same question? @zichufu

[zcfu21]

Sure.

with plt.rc_context({'font.sans-serif': ['Arial']}):
    plt.figure(figsize=(15, 15))
    col_ha = HeatmapAnnotation(GeneNums=anno_barplot(df_bar,
                                           height=10,colors='orangered'),legend=True,legend_gap=5,hgap=0.5)
    row_ha = HeatmapAnnotation(
                            selected=anno_label(label_rows,colors=colors_dict,relpos=(0,0.4)),
                            axis=0,verbose=0,orientation='right'
                            )
    cm = ClusterMapPlotter(data=distance_df,top_annotation=col_ha,right_annotation=row_ha,
                           col_split=6, row_split=6, col_split_gap=1,
                           row_cluster_metric='euclidean',
                           col_cluster_metric='euclidean',
                           row_cluster_method='ward',
                           col_cluster_method='ward',
                           row_split_gap=1,label='distance',
                           row_dendrogram=True,col_dendrogram=False,show_rownames=False,show_colnames=False,
                           tree_kws={'row_cmap': 'Dark2'},cmap='Spectral_r',
                           legend_gap=5,legend_hpad=2,legend_vpad=5)
    ax=col_ha.annotations[0].ax
    ax.yaxis.set_major_locator(MultipleLocator(1)) # set yaxis ticks
    plt.show()

[DingWB]

Looks great.
A minor suggestion:
(1). You can add a parameter extend=True to anno_label, cause I saw the labels in the right annotation overlapped with each other.
Please see here for more details.

(2). You can also choose to set relpos=(0,0.5) to set the endpoint of the curve to the center of the text.

(3). To make it more beautiful, you can also choose to plot the custom box around the cluster:

# plot custom spines
for i in range(cm.heatmap_axes.shape[0]):
    for j in range(cm.heatmap_axes.shape[1]):
        if i != j:
            continue
        ax = cm.heatmap_axes[i][j]
        for side in ["top", "right", "left", "bottom"]:
            ax.spines[side].set_visible(True)
            ax.spines[side].set_color('red')
            ax.spines[side].set_linewidth(2)

See here for example.
(4). I would like to suggest you decrease the figsize, cause the legend is too small and the heatmap looks very big.

[zcfu21]

Thanks for your suggestions! It looks better now.