Usage Principles

Import the Scanpy API as:

import scanpy.api as sc

Workflow

The typical workflow consists of subsequent calls of data analysis tools in sc.tl, e.g.:

sc.tl.tsne(adata, **tool_params)  # embed the data using tSNE

where adata is an AnnData object. Each of these calls adds annotation to an expression matrix X, which stores n_obs observations (cells) of n_vars variables (genes). For each tool, there typically is an associated plotting function in sc.pl:

sc.pl.tsne(adata, **plotting_params)

If you pass show=False, a matplotlib.axes.Axes instance is returned and you have all of matplotlib’s detailed configuration possibilities.

To facilitate writing memory-efficient pipelines, by default, Scanpy tools operate inplace on adata and return None - this also allows to easily transition to out-of-memory pipelines. If you want to return a copy of the AnnData object and leave the passed adata unchanged, pass copy=True.

AnnData

Scanpy is based on anndata, which provides the AnnData class.

At the most basic level, an AnnData object adata stores a data matrix (adata.X), dataframe-like annotation of observations (adata.obs) and variables (adata.var) and unstructured dict-like annotation (adata.uns). Values can be retrieved and appended via adata.obs['key1'] and adata.var['key2']. Names of observations and variables can be accessed via adata.obs_names and adata.var_names, respectively. AnnData objects can be sliced like dataframes, for example, adata_subset = adata[:, list_of_gene_names]. For more, see this blog post.

To read a data file to an AnnData object, call:

adata = sc.read(filename)

to initialize an AnnData object. Possibly add further annotation using, e.g., pd.read_csv:

import pandas as pd
anno = pd.read_csv(filename_sample_annotation)
adata.obs['cell_groups'] = anno['cell_groups']  # categorical annotation of type pandas.Categorical
adata.obs['time'] = anno['time']                # numerical annotation of type float
# alternatively, you could also set the whole dataframe
# adata.obs = anno

To write, use:

adata.write(filename)
adata.write_csvs(filename)
adata.write_loom(filename)