Plotting functions

Plotting functions in PIASO

This notebook will demonstrate the different plotting functions and their application to sample data.

import piaso

import piaso
import numpy as np
import matplotlib.pyplot as plt

piaso.settings.set_figure_params(style="cell")

Every plot on this page is a piaso.pl call. set_figure_params fixes the figure size, font and the rest of the house style, so the rcParams block that used to follow it is unnecessary — and scanpy is not imported, because nothing here needs it.

style="cell" targets Cell’s single-column figure spec: 7 pt type on a 3.35-inch (85 mm) canvas. That is why titles and legends look small when you draw at figsize=(16, 8) — the type stays 7 pt while the canvas grows. Either draw at the size the style is for, or pass fontsize= to set_figure_params when you deliberately want a larger figure.

warnings.simplefilter(action='ignore', category=FutureWarning)

Load the data

Let us load a 20k subsampled version of the Seattle Alzheimer’s Disease Brain Cell Atlas (SEA-AD) project dataset described in detail in Gabitto et. al. (2024). We will be using the scRNA-seq data from the dataset in this tutorial.

Download the subsampled dataset from Google Drive: https://drive.google.com/file/d/1EdRA0ECvPlEnNaOzKqj19GrEtucB7hmE/view?usp=drive_link

The original data is available on https://portal.brain-map.org/explore/seattle-alzheimers-disease.

adata = piaso.data.load_dataset("sea_ad_mtg_20k")
adata

Visualize with a discrete color map

The discrete color map can be used to visualize the categorical variables such as cell subclass.

piaso.pl.embedding(adata,
                   basis='X_umap',
                   color=['Subclass'],
                   palette=piaso.pl.color.d_color3,
                   legend_fontoutline=2)

piaso.pl.embedding(adata,
                   basis='X_umap',
                   color=['Subclass'],
                   palette=piaso.pl.color.d_color3,
                   legend_fontoutline=2,
                   legend_fontsize=7,
                   legend_loc='on_data',
                   legend_fontbackground='white')

Visualize with a continuous color map

The continuous color map can be used to visualize the continuous variables such as gene expression.

piaso.pl.umap(adata,
              color=['GAD1', 'PVALB', 'SST', 'PDGFRA'],
              cmap=piaso.pl.color.c_color1,
              ncols=2)

piaso.pl.umap(adata,
              color=['GAD1', 'PVALB', 'SST', 'PDGFRA'],
              cmap=piaso.pl.color.c_color4,
              ncols=2)

Split the UMAP by condition

mapping_dict=dict(zip(adata.obs['CERAD score'], adata.obs['CERAD score']))

mapping_dict={'Absent': 'Absent',
 'Sparse': 'Disease',
 'Moderate': 'Disease',
 'Frequent': 'Disease'}

adata.obs['Condition']=adata.obs['CERAD score'].map(mapping_dict)

piaso.pl.embedding(adata,
                   basis='X_umap',
                   color=['Condition'],
                   palette=piaso.pl.color.d_color1,
                   legend_fontoutline=2)

Visualizing the expression of a specific gene across two different experimental conditions

piaso.pl.plot_embeddings_split(adata,
                               color='ADGRG7',
                               layer=None,
                               splitby='Condition',
                               color_map=piaso.pl.color.c_color1,
                               size=100,
                               frameon=False,
                               legend_loc=None,)

piaso.pl.plot_embeddings_split(adata,
                               color='FEZF2',
                               layer=None,
                               splitby='Condition',
                               color_map=piaso.pl.color.c_color1,
                               size=100,
                               frameon=False,
                               legend_loc=None,)

Visulaizing continuous variables (Number of UMIs) across different experimental conditions

piaso.pl.plot_embeddings_split(adata,
                               color='Number of UMIs',
                               layer=None,
                               splitby='Condition',
                               color_map=piaso.pl.color.c_color4,
                               size=100,
                               frameon=False,
                               legend_loc=None)

Visualizing categorical variables (Cell subclass) across different experimental conditions

piaso.pl.plot_embeddings_split(adata,
                               color='Subclass',
                               layer=None,
                               splitby='Condition',
                               size=100,
                               frameon=False,)

We can visualize these cell subclasses alongside the subclass legend. To enhance clarity, the ncol parameter can be used to adjust the number of subplots per row.

piaso.pl.plot_embeddings_split(adata,
                               color='Subclass',
                               layer=None,
                               splitby='Condition',
                               size=100,
                               frameon=False,
                               ncol=1)

piaso.pl.plot_embeddings_split(adata,
                               color='Subclass',
                               layer=None,
                               splitby='Condition',
                               size=100,
                               frameon=False,
                               legend_fontsize=7,
                               legend_fontoutline=2,
                               legend_fontweight=5,
                               legend_loc='on data')

Violin plots of grouped data

Visulaize how the groups in the dataset vary across a list of different features. Here, you can plot a different subplot for each feature of interest.

piaso.pl.plot_features_violin(adata,
                              feature_list=[ 'CALB2', 'GAD2', 'SATB2', 'TBR1', 'Number of UMIs'],
                              width_single=8,
                              height_single=2.3,
                              groupby='Subclass',
                              show_grid=False)

Adding the horizontal grid line

piaso.pl.plot_features_violin(adata,
                              feature_list=[ 'CALB2', 'GAD2', 'SATB2', 'TBR1', 'Number of UMIs'],
                              width_single=8,
                              height_single=2.3,
                              groupby='Subclass')

The height and widhts of these plots can be adjusted according to the number of features and groups you have for better visualization.

piaso.pl.plot_features_violin(adata,
                              feature_list=[ 'CALB2', 'GAD2', 'SATB2', 'TBR1', 'Number of UMIs'],
                              width_single=20,
                              height_single=3,
                              groupby='Subclass',
                              show_grid=False)

Violin plots of entire data

We can visualize the entire dataset as a single group, adjusting the dot size to plot all the cells onto the violin plot.

piaso.pl.plot_features_violin(adata,
                              feature_list=[ 'Genes detected'],
                              width_single=4,
                              height_single=2.0,
                              size=1)

We can visualize the distributions of the cells across various features together.

piaso.pl.plot_features_violin(adata,
                              feature_list=[ 'MALAT1', 'GAD2', 'Number of UMIs'],
                              width_single=4,
                              height_single=2.3,
                              size=0.5,
                              show_grid=False)

Save output plots

Use the save parameter to specific the file name and path to save:

piaso.pl.plot_embeddings_split(adata,
                               color='ADGRG7',
                               layer=None,
                               splitby='Condition',
                               color_map=piaso.pl.color.c_color1,
                               size=100,
                               frameon=False,
                               legend_loc=None,
                               save='./PIASO_UMAP_split_by_condition.pdf')

piaso.pl.plot_features_violin(adata,
                              feature_list=[ 'CALB2', 'GAD2', 'SATB2', 'TBR1', 'Number of UMIs'],
                              width_single=8,
                              height_single=2.3,
                              groupby='Subclass',
                              show_grid=False,
                              save='./violin_plot_piaso.pdf')

Combine categorical variables

mapping_dict={'Absent': 'Control',
 'Sparse': 'Disease',
 'Moderate': 'Disease',
 'Frequent': 'Disease'}

adata.obs['Condition']=adata.obs['CERAD score'].map(mapping_dict)

adata.obs['SubclassXCondition'] = piaso.pp.getCrossCategories(adata.obs, 'Subclass', 'Condition', )

By combining multiple categories, while maintaining the original orders of the categories, you could show the gene expression changes at cell type level across different conditions.

piaso.pl.dotplot(
    adata,
    features=['SERPINE1'],
    groupby='SubclassXCondition',
    swap_axes=True,
    cmap=piaso.pl.color.c_color4
)

We can also switch the order in which we combine the categorical variables.

adata.obs['ConditionXSubclass'] = piaso.pp.getCrossCategories(adata.obs, 'Condition', 'Subclass',)

piaso.pl.dotplot(
    adata,
    features=['SERPINE1'],
    groupby='ConditionXSubclass',
    swap_axes=True,
    cmap=piaso.pl.color.c_color4
)

Count the number of cells in each category

piaso.pp.table(adata.obs['Subclass'])

You can use the rank parameter to order the results, and the ascending parameter to determine the order of the ranking.

piaso.pp.table(adata.obs['Subclass'], rank = True, ascending = True)

piaso.pp.table(adata.obs['Subclass'], rank = True, ascending = False)

In order to store the results as a dataframe, set as_dataframe = True.

subclass_df = piaso.pp.table(adata.obs['Subclass'], rank = True, ascending = False, as_dataframe = True)

Category order and colours

Legends and dot-plot rows follow the category order of the grouping column, and colours come with it. Both backends let you set these once and have every later plot honour them — the difference is where the setting lives.

AnnData: order lives in the categorical dtype, colours in adata.uns['<column>_colors'] (one hex colour per category, in category order):

subclass_order = ['Astro', 'Oligo', 'OPC', 'Micro', 'Endo']  # your order
adata.obs['Subclass'] = adata.obs['Subclass'].cat.reorder_categories(subclass_order)
adata.uns['Subclass_colors'] = ['#4c72b0', '#dd8452', '#55a868', '#c44e52', '#8172b3']

piaso.pl.plotEmbedding(adata, color='Subclass')  # legend now uses your order + palette

Note these live in the Python object: they persist only if you write the AnnData back to disk afterwards.

Cytome: set_categories persists order and palette in the file, so they apply in every later session without re-declaring them:

ds.set_categories(
    'Subclass',
    order=subclass_order,
    colors=dict(zip(subclass_order,
                    ['#4c72b0', '#dd8452', '#55a868', '#c44e52', '#8172b3'])),
)

piaso.pl.plotEmbedding(ds, color='Subclass')  # same legend — today and next week

Set once, every plot afterwards honours it, including in a fresh session — the ordering travels with the data rather than with your script.