Ctrl+K

Copy Number Heatmaps

Copy Number Heatmaps#

scgenome can be used to plot a heatmap of copy number changes across the genome for a set of cells. The heatmaps can be annotated based on per cell properties stored in the obs dataframe of the AnnData.


import scgenome
import matplotlib.pyplot as plt

adata = scgenome.datasets.OV2295_HMMCopy_reduced()

Basic Heatmaps#

The scgenome.pl.plot_cell_cn_matrix function can be used to produce a basic heatmap plot of copy number with rows representing cells (or clones) and columns bins of the genome. Use the layer_name arg to specify the layer to plot. By default the standard copy number palette will be used to map integer copy number values to heatmap colors.


g = scgenome.pl.plot_cell_cn_matrix(adata, layer_name='state')
scgenome.pl.cn_legend(plt.gca())

Plotting raw values is also possible with raw=True. Use layer_name=None to do a heatmap of .X. The vmin and vmax parameters can be used to anchor the colormapping.


g = scgenome.pl.plot_cell_cn_matrix(adata, layer_name='copy', raw=True, vmin=0, vmax=4)

Specifying Cell Order#

Cells can be ordered in a heatmap by specifying any set of one or more fields from .obs. A legend can be added to a specific matplotlib axis, either the heatmap axis or a distinct axis to allow for flexible positioning of the legend.


g = scgenome.pl.plot_cell_cn_matrix(
    adata,
    cell_order_fields=['cell_order'])
scgenome.pl.cn_legend(plt.gca())

Annotated Heatmaps#

The scgenome.pl.plot_cell_cn_matrix_fig function can be used to produce more complex heatmaps. Cells can annotated with any field from .obs using annotation_fields or .var using var_annotation_fields. Categorical fields will be given a discrete color map. Real valued fields will be given a continuous color map.


g = scgenome.pl.plot_cell_cn_matrix_fig(
    adata[:, adata.var['gc'] > 0],
    cell_order_fields=['cell_order'],
    annotation_fields=['cluster_id', 'sample_id', 'quality'],
    var_annotation_fields=['gc'],
)

Cytoband staining#

The scgenome.tl.add_cyto_giemsa_stain function can be used to calculate per bin cytoband stain, which can be added to a complex heatmap by adding 'cyto_band_giemsa_stain' to var_annotation_fields.


adata.var = scgenome.tl.add_cyto_giemsa_stain(adata.var)

g = scgenome.pl.plot_cell_cn_matrix_fig(
    adata[:, adata.var['chr'] == '1'],
    layer_name='copy',
    cell_order_fields=['cell_order'],
    annotation_fields=['cluster_id', 'sample_id'],
    var_annotation_fields=['gc', 'cyto_band_giemsa_stain'],
    raw=True, vmin=0, vmax=4,
)

Irregular bin widths#

Irregular bin widths will result in an irregular heatmap. The scgenome.tl.rebin_regular function can be used to rebin data into a consistent bin widths before generating a heatmap. The heatmap below does not have consistent bin widths resulting in segments and chromosomes having incorrect relative sizes.


adata_irregular = scgenome.datasets.OV_051_Medicc2_reduced()

plt.figure()
g = scgenome.pl.plot_cell_cn_matrix(adata_irregular, layer_name=None)

An AnnData holding copy number data can be rescaled with scgenome.tl.rebin_regular as shown below. Specify the new bin width and the functions for aggregating X, layers and var. Set outer_join=True to ensure all bins are represented in the new var, with NaN where there were no overlapping bins in the var of the input AnnData. Otherwise the new var is subset to only regions overlapping the var of the input AnnData.


adata_regular = scgenome.tl.rebin_regular(
    adata_irregular, 500000, outer_join=True, agg_X=scgenome.tl.bin_width_weighted_mean, agg_layers={},
    agg_var={'is_normal': ('is_normal', scgenome.tl.bin_width_weighted_mean)})

g = scgenome.pl.plot_cell_cn_matrix(adata_regular, layer_name=None)
Contents