Reproduce the heatmap from inferCNV

This document demonstrates to reproduce how the example heatmap from the original R inverCNV implementation. It is based on a small, 183-cell example dataset of malignant and non-malignant cells from Oligodendroglioma derived from Tirosh et al. (2016).

import infercnvpy as cnv
import scanpy as sc

import warnings

warnings.simplefilter("ignore")

Prepare and inspect dataset

The example dataset is available in the datasets module. It is already TPM-normalized, but not log-transformed.

adata = cnv.datasets.oligodendroglioma()
sc.pp.log1p(adata)

It also already has the genomic positions annotated inadata.var:

adata.var.head()

	chromosome	start	end	n_counts	n_cells
gene_symbol
WASH7P	chr1	14363	29806	534.271301	116
LINC00115	chr1	761586	762902	203.649094	18
NOC2L	chr1	879584	894689	880.449707	63
SDF4	chr1	1152288	1167411	1013.251404	70
UBE2J2	chr1	1189289	1209265	411.488953	33

It contains four types of malignant cells, and two clusters of non-malignant cells.

sc.tl.pca(adata)
sc.pp.neighbors(adata)
sc.tl.umap(adata)
sc.pl.umap(adata, color="cell_type")

Run infercnvpy

In this case we know which cells are non-malignant. For best results, it is recommended to use the non-malignant cells as a background. We can provide this information using reference_key and reference_cat.

In order to reproduce the results as exactely as possible, we use a window_size of 100 and a step of 1.

%%time
cnv.tl.infercnv(
    adata,
    reference_key="cell_type",
    reference_cat=["Oligodendrocytes (non-malignant)", "Microglia/Macrophage"],
    window_size=100,
    step=1,
)

CPU times: user 85 ms, sys: 169 ms, total: 254 ms
Wall time: 462 ms

%%time
cnv.pl.chromosome_heatmap(adata, groupby="cell_type", dendrogram=True)

WARNING: dendrogram data not found (using key=dendrogram_cell_type). Running `sc.tl.dendrogram` with default parameters. For fine tuning it is recommended to run `sc.tl.dendrogram` independently.
WARNING: You’re trying to run this on 9013 dimensions of `.X`, if you really want this, set `use_rep='X'`.
         Falling back to preprocessing with `sc.pp.pca` and default params.
WARNING: Groups are not reordered because the `groupby` categories and the `var_group_labels` are different.
categories: Microglia/Macrophage, Oligodendrocytes (non-malignant), malignant_93, etc.
var_group_labels: chr1, chr2, chr3, etc.
CPU times: user 3.1 s, sys: 6.1 s, total: 9.2 s
Wall time: 1.52 s

Note that running the same analysis in R (invercnv v1.6.0 from Bioconductor) takes about 1:30 min.