Skip to content

Latest commit

 

History

History
158 lines (106 loc) · 5.35 KB

README.md

File metadata and controls

158 lines (106 loc) · 5.35 KB
Raw

pseudo-RNAscope

pseudo_rnascope is a simple package to plot gene expression of multiple genes as different colors in the same plot using any scanpy plotting function. The original use-case is to mimic RNAscope with visium data.

pseudo_rnascope_limb visualisation of a developing limb with data from: B. Zhang et al. (2023) ‘A human embryonic limb cell atlas resolved in space and time.’ Nature 2023. DOI: 10.1038/s41586-023-00000-0.

TLDR

Run like this...

To plot the expression overlap of two genes, e.g. FGF2 and FGFR2, in spatial locations of visium data

channel1 = "FGF2"
channel2 = "FGFR2"

run add_pseudo_rna_scope on an existing anndata object:

import scanpy as sc
import matplotlib as mpl
import matplotlib.pyplot as plt
from pseudo_rnascope import add_pseudo_rna_scope

ranges = add_pseudo_rna_scope(
    adata,

    channel1 = channel1,
    channel1_color = (1,0,0),  # red (RGB)

    channel2 = channel2,
    channel2_color = (0,1,0),  # green (RGB)

    auto_range_quantiles = (0.2, 0.9),
    knn_smooth = True,
    gamma = 10,
)

This will use red for channel1 (FGF2) and green for channel2 (FGFR2), hence overlaps will appear yellow.

  • Dynamic ranges can be set for both channels explicitly, or automatically via upper and lower gene expression quantiles with auto_range_quantiles.
  • Neareast-neighbor smoothing of expression values can be added to mitigate data sparsity with knn_smooth=True.
  • Larger gamma values make mixed colors appear brighter, while gamma=1 corresponds to linear color mixing.

Scanpy plotting functions can then be used by specifying the added adata.obs column "pseudo_RNAscope".

sc.set_figure_params(figsize=[16,16], dpi=75)

sc.pl.spatial(
    adata, 
    img_key="hires", 
    color='pseudo_RNAscope', 
    size=1.5,
    legend_loc=None,  # turn off legend (necessary)
    alpha = adata.obs["pseudo_RNAscope_alpha"],
    show=False,
)

plt.colorbar(mpl.cm.ScalarMappable(norm=mpl.colors.Normalize(ranges[channel2]["vmin"], ranges[channel2]["vmax"]), cmap='Greens'),
             orientation='vertical', label=channel2, extend='both', shrink=0.5, pad=-0.04)

plt.colorbar(mpl.cm.ScalarMappable(norm=mpl.colors.Normalize(ranges[channel1]["vmin"], ranges[channel1]["vmax"]), cmap='Reds'),
             orientation='vertical', label=channel1, extend='both', shrink=0.5, pad=0.03)

Corresponding colors for all spots are saved in adata.uns["pseudo_RNAscope_colors"] and another anndata.obs column called "pseudo_RNAscope_alpha" allows to set alpha values, making spots with low color intensity transparent.

Computed colors are also encoded as decimal numbers between 0 and 1, and stored in adata.obs["pseudo_RNAscope_alt"]. This can be used as an alternative option for plotting (also works with squidpy.pl.spatial_scatter()).

alternative

Use values stored in adata.obs["pseudo_RNAscope_alt"] and pass adata.uns["pseudo_RNAscope"]["cmap"] as a colormap to decode them:

sc.set_figure_params(figsize=[16,16], dpi=75)

sc.pl.spatial(
    adata, 
    img_key="hires", 
    color='pseudo_RNAscope_alt', 
    size=1.5,
    cmap = adata.uns['pseudo_RNAscope']['cmap'],
    vmin=0, # use all values (necessary)
    vmax=1, # use all values (necessary)
    colorbar_loc=None,  # turn off native colorbar, we need separate ones per channel
    alpha = adata.obs.sort_values("pseudo_RNAscope_alt")["pseudo_RNAscope_alpha"],  # since values will be sorted before plotting
    show=False,
)

plt.colorbar(mpl.cm.ScalarMappable(norm=mpl.colors.Normalize(ranges[channel2]["vmin"], ranges[channel2]["vmax"]), cmap='Greens'),
             orientation='vertical', label=channel2, extend='both', shrink=0.5, pad=-0.04)

plt.colorbar(mpl.cm.ScalarMappable(norm=mpl.colors.Normalize(ranges[channel1]["vmin"], ranges[channel1]["vmax"]), cmap='Reds'),
             orientation='vertical', label=channel1, extend='both', shrink=0.5, pad=0.03)

System requirements

show requirements

Hardware requirements

pseudo_rnascope can run on a standard computer with enough RAM to hold the used datasets in memory.

Software requirements

OS requirements

The package has been tested on:

  • macOS Monterey (12.6.7)
  • Linux: Ubuntu 18.04.6 bionic

Python requirements

A python version >=3.0 is required. Various python libraries are used, listed in pyproject.toml. Currently only the python scientific stack (numpy, scipy) is required as a dependency, however, it only works together with an anndata object from the scanpy package as input and for downstream plotting. pseudo_rnascope and all dependencies can be installed via pip (see below).

Installation

Install with pip

from github

pip install git+https://github.com/JPatrickPett/pseudo_rnascope.git

from PyPI

will be added

(installation time: <1 min)

Usage and Documentation

Please refer to the demo notebook. There's a function docstring in the source code, which will be rendered on ReadTheDocs once the package goes live.

(demo running time: around 2 min)

Citation

pseudo_rnascope is part of the forthcoming manuscript "A multiomic atlas of human early skeletal development" by To, Fei, Pett et al. Stay tuned for details!