Getting started

Getting started#

This page uses a tiny simulated AnnData. Swap it for your own object.

Install#

pip install -e ".[dev]"

Load and check#

import numpy as np
from anndata import AnnData
import cellfish as cf

rng = np.random.default_rng(0)
adata = AnnData(rng.normal(size=(80, 20)))
adata.obs_names = [f"c{i}" for i in range(80)]
adata.var_names = [f"g{i}" for i in range(20)]
adata.obs["cell_type"] = (["Hepatocyte"] * 40) + (["Macrophage"] * 40)
adata.obs["sample"] = (["s1"] * 20 + ["s2"] * 20) * 2
adata.obsm["X_umap"] = rng.normal(size=(80, 2))
adata.obsm["spatial"] = rng.uniform(0, 10, size=(80, 2))

cf.data.require_obs(adata, ["cell_type", "sample"])
cf.data.require_obsm(adata, "umap")  # also accepts X_umap

Style and palettes#

Paper palettes stay in the analysis repo. Pass a dict:

MY_PALETTE = {"Hepatocyte": "#1F577B", "Macrophage": "#E069A6"}

cf.pl.setup_style()
cf.pl.reorder_and_set_palettes(adata, "cell_type", palette=MY_PALETTE)

Embedding (UMAP or tissue)#

The same function draws both. Change basis=.

ax = cf.pl.embedding(adata, basis="X_umap", color="cell_type", show=False)
cf.pl.add_contour(ax, adata, groupby="cell_type", clusters=["Hepatocyte"], basis="X_umap")

cf.pl.embedding(adata, basis="spatial", color="cell_type", show=False)

cf.pl.umap(...) is a shortcut that requires adata.obsm["X_umap"].

Composition#

props = cf.pl.get_cluster_proportions(
    adata, cluster_key="cell_type", sample_key="sample"
)
fig = cf.pl.plot_cluster_proportions(
    props, cluster_palette=MY_PALETTE, show=False
)
cf.pl.savefig(fig, "figures/cell_type_proportions.pdf")

Write AnnData#

cf.io.write_h5_safe(adata, "adata.clean.h5ad")

Next#