pynebulosa 0.1.0

Single-cell data visualization using kernel gene-weighted density estimation

pyNebulosa

Single-cell data visualization using kernel gene-weighted density estimation.

pyNebulosa is a Python package (ported from the original R/Bioconductor package) that recovers gene expression signals lost to dropout in single-cell data by using weighted kernel density estimation on low-dimensional embeddings. It integrates natively with scanpy and AnnData.

Motivation

Standard feature plots in single-cell analysis are plagued by sparsity — many cells show zero expression even for marker genes, making it hard to see biological signal. pyNebulosa addresses this by computing a gene-weighted kernel density estimate (KDE) on the embedding space, effectively smoothing expression over neighbouring cells.

Standard feature plot	pyNebulosa density plot
Sparse, hard to interpret	Smooth, recovers signal

R vs Python

pyNebulosa was originally an R/Bioconductor package (Nebulosa) for Seurat and SingleCellExperiment objects. This Python port provides an equivalent API for the scanpy/AnnData ecosystem, producing identical density estimates.

R (Seurat)	Python (scanpy)
library(Nebulosa) library(Seurat) pbmc <- pbmc3k.SeuratData::pbmc3k.final plot_density(pbmc, "NKG7") plot_density(pbmc, c("NKG7", "GNLY"), joint = TRUE)	import scanpy as sc import pynebulosa as nb adata = sc.datasets.pbmc3k_processed() nb.plot_density(adata, "NKG7") nb.plot_density(adata, ["NKG7", "GNLY"], joint=True)

Installation

pip install pynebulosa

With scanpy (recommended):

pip install "pynebulosa[scanpy]"

Quick start

import scanpy as sc
import pynebulosa as nb

adata = sc.datasets.pbmc3k_processed()

# Single gene density plot
nb.plot_density(adata, "NKG7")

Multi-marker panels

Visualize multiple markers at once to identify cell populations:

nb.plot_density(adata, ["MS4A1", "NKG7", "CST3", "PPBP"], ncols=4)

Joint density

Find co-expressing cell populations by computing the product of individual gene densities:

nb.plot_density(adata, ["NKG7", "GNLY"], joint=True)

API

Function	Purpose
nb.plot_density(adata, features)	Main visualization function
nb.calculate_density(weights, coords)	Compute density values programmatically
nb.wkde2d(x, y, w)	Low-level weighted 2D KDE

Key parameters

Parameter	Description	Default
joint	Compute joint density across features	False
reduction	Embedding key in adata.obsm	Auto-detect (X_umap)
method	"wkde" (custom) or "ks" (scipy)	"wkde"
adjust	Bandwidth adjustment factor	1.0
cmap	Matplotlib colormap	"viridis"
layer	Expression layer	adata.X

How it differs from similar tools

Tool	Approach	Use case
scanpy.tl.embedding_density	Unweighted KDE per category	Where are cells of group X?
pyUCell	Rank-based scoring + KNN smoothing	Gene signature scoring
pyNebulosa	Gene-weighted KDE on embeddings	Recovering dropout signal

Documentation

• Full documentation (ReadTheDocs)
• Tutorial notebook (PBMC 3K walkthrough)

Citation

If you use pyNebulosa, please cite:

Alquicira-Hernandez, J., Powell, J.E. Nebulosa recovers single-cell gene expression signals by kernel density estimation. Bioinformatics, 37(16), 2485-2487, 2021. doi:10.1093/bioinformatics/btab003