Installation and quickstart

Supported Python

easydecon requires Python 3.10 or newer. The package metadata currently advertises support for Python 3.10 through 3.13.

Installation

Install the release package from PyPI:

python -m pip install easydecon

For development from a checkout:

python -m pip install -e ".[test]"

Install optional extras only when needed:

python -m pip install -e ".[spatial]"
python -m pip install -e ".[deseq]"
python -m pip install -e ".[docs]"
python -m pip install -e ".[test]"

The spatial extra is needed for SpatialData containers and related plotting/query helpers. The deseq extra is needed for marker_method="pydeseq2". Core AnnData workflows use the required package dependencies.

Minimal input expectations

You need:

an AnnData table, or a SpatialData object containing an AnnData-like table;
an expression matrix whose var_names are gene identifiers;
a marker table with at least group and names columns; and
marker gene identifiers that overlap the spatial expression var_names.

The generic term in the docs is “spatial location”. A spatial location may be a spot, bin, or segmented cell depending on the upstream data.

Minimal workflow

import easydecon as ed

result = ed.run_easydecon(
    sdata=sdata,
    markers_df=markers_df,
    return_result_object=True,
    verbose=False,
)

result.posterior_df
result.assigned_labels
result.diagnostics

posterior_df contains relative posterior support among tested marker groups, not guaranteed absolute biological cell fractions. assigned_labels contains hard assignments and therefore discards uncertainty. Inspect diagnostics before relying on assignments downstream.

run_easydecon defaults to the standard Phase 1 permutation workflow and the default Phase 2 weighted Jaccard method.

Fast exploratory run

Quantile filtering is a fast exploratory shortcut. The standard Phase 1 workflow uses permutation filtering, so final analyses should normally return to filtering_algorithm="permutation".

result = ed.run_easydecon(
    sdata=sdata,
    markers_df=markers_df,
    filtering_algorithm="quantile",
    method="wjaccard",
    return_result_object=True,
    verbose=False,
)

Next steps

Marker inputs and table schema explains marker tables, generated markers, and PreparedMarkers.
How easydecon works explains the full Phase 1 -> Phase 2 -> posterior flow.
Understanding EasyDeconResult explains which matrix to use for each downstream task.
Visualizing easydecon results gives Matplotlib recipes for maps and summaries.

Compatibility tuple return

Without return_result_object=True, run_easydecon returns the historical five-value tuple:

phase1_result, phase2_result, assigned_labels, priors_df, assignment_df = ed.run_easydecon(
    sdata=sdata,
    markers_df=markers_df,
)

Set return_diagnostics=True to append the diagnostics dictionary to that tuple. New code should prefer return_result_object=True because it exposes likelihoods_df, posterior_df, assignment_df, and marker diagnostics with stable attribute names.