GitHub - MohammadFaizIqbalFaiz/spartan-st: A python toolkit for spatial domains and spatially variable genes discovery for high-resolution spatially resolved transcriptomics data.

Conceptual overview of Spartan for spatial domain identification and spatially variable gene discovery.

Spartan

Spartan is an activation-aware spatial transcriptomics framework for spatial domain identification and spatially variable gene discovery.

Spartan integrates spatial topology, gene expression connectivity, and Local Spatial Activation (LSA) into an aggregated graph for unsupervised spatial domain detection. The framework is designed for multiple spatial transcriptomics technologies, including imaging-based datasets such as MERFISH, sequencing-based datasets such as Stereo-seq, and high-resolution platforms such as Visium HD. Spartan can operate on both classic AnnData and next-generation spatial omics, SpatialData frameworks.

Overview

Spatial transcriptomics datasets contain both molecular and spatial information. Many spatial clustering approaches primarily model local similarity or spatial smoothing. Spartan instead introduces a Local Spatial Activation graph that captures neighborhood-conditioned transcriptional deviation across spatial neighborhoods.

Spartan constructs three complementary graphs:

Spatial graph (S) — captures physical neighborhood topology.
Gene expression connectivity graph (G) — captures transcriptomic similarity in reduced expression space.
Local Spatial Activation graph (L) — captures local transcriptional activation/deviation structure across spatial neighborhoods.

These graphs are combined into an aggregated graph:

$$J = (\alpha-\beta_1)L + (1-\alpha)G + (\alpha-\beta_2)S$$

Leiden clustering is then applied to the aggregated graph to identify spatial domains.

Spartan also provides a Spatial Activation Quotient (SAQ) workflow for identifying spatially variable genes using the LSA graph.

Installation

General package environment

For general users, create the lightweight Spartan environment:

conda env create -f envs/environment.core.yml
conda activate spartan-core
pip install -e .

Paper reproducibility environment

For exact reproduction of manuscript analyses and tutorial notebooks:

conda env create -f envs/environment.paper.lock.yml
conda activate spartan-paperS
pip install -e .

The paper environment contains the pinned package versions used for the manuscript analyses.

Quickstart tutorial

This quickstart is intended for new Spartan users. It demonstrates a minimal end-to-end workflow using the representative MERFISH dataset available from Squidpy. The workflow loads one MERFISH section, preprocesses the data, identifies Spartan spatial domains, runs SAQ-based spatially variable gene discovery, and visualizes both spatial domains and representative SVGs.

1. Import packages and load a MERFISH sample

import spartan as sp
import squidpy as sq
import scanpy as sc
import matplotlib.pyplot as plt

adata_data = sq.datasets.merfish()  # representative MERFISH dataset from Squidpy

# Extract one sample/section
adata = adata_data[adata_data.obs["Bregma"] == 1].copy()

2. Add a ground-truth column if needed

Some Spartan workflows and tutorials expect a ground_truth column for downstream benchmarking or visualization. The core Spartan domain workflow itself does not require ground-truth labels.

# If a ground-truth column is absent, initialize one
adata.obs["ground_truth"] = ""

3. Preprocess the imaging-based dataset

For MERFISH and related imaging-based spatial transcriptomics datasets, use the imaging preprocessing wrapper. To use this wrapper, ground_truth column is needed, create one using code snippet in 2.

adata = sp.tl.pre_process_imaging(adata)

4. Run Spartan spatial domain identification

This step constructs the spatial graph, Local Spatial Activation graph, gene-expression connectivity graph, aggregated Spartan graph, and then applies Leiden clustering to identify spatial domains.

sp.tl.spartan_spatial_domains(
    adata,
    spatial_coord="generic",
    spatial_neighborhood="knn",
    spatial_neighs=10,
    gene_neighs=15,
    alpha=0.69,
    beta1=0.26,
    beta2=0.24,
    resolution=0.61,
    seed=1,
    key_added="spartan_domains",
)

The predicted spatial domains are stored in:

adata.obs["spartan_domains"]

Inspect the number of cells assigned to each Spartan domain:

adata.obs["spartan_domains"].value_counts()

5. Run SAQ-based spatially variable gene discovery

Spartan uses the Local Spatial Activation graph to compute Spatial Activation Quotient scores for spatially variable gene discovery.

sp.tl.spartan_svg(
    adata,
    lsa_graph=adata.obsp["spartan_lsa_graph"],
    top_k_refine=161,
    key_added="spartan_svg",
)

View the top SAQ-ranked genes:

adata.var.sort_values("spartan_saq", ascending=False).head(10)

6. Visualize cell classes and Spartan spatial domains

sc.pl.spatial(
    adata,
    img_key=None,
    color=["Cell_class", "spartan_domains"],
    wspace=0.4,
    title=["Cell Class", "Spatial Domains"],
    size=2.5,
    spot_size=0.005,
)

Example output:

7. Visualize representative Spartan SVGs

genes = ["Ucn3", "Mbp", "Sln", "Nnat"]

sc.pl.spatial(
    adata,
    img_key=None,
    color=genes,
    color_map="Reds",
    size=2.5,
    spot_size=0.005,
    legend_loc=None,
)

Example output:

8. Main objects created by the quickstart

After running the quickstart, Spartan stores the main outputs inside the AnnData object.

Spatial domain labels:

adata.obs["spartan_domains"]

Graph outputs:

adata.obsp["spartan_spatial_graph"]
adata.obsp["spartan_spatial_weights"]
adata.obsp["spartan_lsa_graph"]
adata.obsp["spartan_gene_graph"]
adata.obsp["spartan_joint_graph"]

SAQ/SVG outputs:

adata.var["spartan_saq"]
adata.var["spartan_saq_pval"]
adata.var["spartan_saq_fdr"]
adata.var["spartan_svg"]
adata.var["spartan_saq_rank"]

Complete quickstart code

import spartan as sp
import squidpy as sq
import scanpy as sc
import matplotlib.pyplot as plt

adata_data = sq.datasets.merfish()
adata = adata_data[adata_data.obs["Bregma"] == 1].copy()

adata.obs["ground_truth"] = ""

adata = sp.tl.pre_process_imaging(adata)

sp.tl.spartan_spatial_domains(
    adata,
    spatial_coord="generic",
    spatial_neighborhood="knn",
    spatial_neighs=10,
    gene_neighs=15,
    alpha=0.69,
    beta1=0.26,
    beta2=0.24,
    resolution=0.61,
    seed=1,
    key_added="spartan_domains",
)

adata.obs["spartan_domains"].value_counts()

sp.tl.spartan_svg(
    adata,
    lsa_graph=adata.obsp["spartan_lsa_graph"],
    top_k_refine=161,
    key_added="spartan_svg",
)

adata.var.sort_values("spartan_saq", ascending=False).head(10)

sc.pl.spatial(
    adata,
    img_key=None,
    color=["Cell_class", "spartan_domains"],
    wspace=0.4,
    title=["Cell Class", "Spatial Domains"],
    size=2.5,
    spot_size=0.005,
)

genes = ["Ucn3", "Mbp", "Sln", "Nnat"]

sc.pl.spatial(
    adata,
    img_key=None,
    color=genes,
    color_map="Reds",
    size=2.5,
    spot_size=0.005,
    legend_loc=None,
)

For sequencing-based spatial transcriptomics, use:

adata = sp.tl.pre_process_sequencing(adata)

Main outputs

Spatial domains

Spartan stores predicted spatial domain labels in:

adata.obs["spartan_domains"]

or in the user-defined column specified by key_added.

Graphs stored in `adata.obsp`

Spartan stores the graph components used for spatial domain detection:

Key	Description
`spartan_spatial_graph`	Spatial neighborhood adjacency graph
`spartan_spatial_weights`	Row-normalized spatial weight matrix
`spartan_lsa_graph`	Local Spatial Activation graph
`spartan_gene_graph`	Gene expression connectivity graph
`spartan_joint_graph`	Aggregated graph used for Leiden clustering

Spatially variable gene outputs

The SAQ/SVG workflow stores gene-level statistics in adata.var:

Key	Description
`spartan_saq`	Spatial Activation Quotient score
`spartan_saq_pval`	SAQ p-value
`spartan_saq_fdr`	Benjamini–Hochberg adjusted FDR
`spartan_svg`	Boolean SVG calls
`spartan_saq_rank`	Gene ranking by SAQ score

Core user-facing API

Preprocessing

`pre_process_imaging`

adata = sp.tl.pre_process_imaging(adata)

Preprocesses imaging-based spatial transcriptomics datasets such as MERFISH and Vizgen MERFISH.

This workflow is intended for single-cell imaging datasets, where measured genes are typically retained and standard normalization/log transformation is applied before graph construction.

`pre_process_sequencing`

adata = sp.tl.pre_process_sequencing(adata)

Preprocesses sequencing-based spatial transcriptomics datasets such as Stereo-seq, Visium, and Visium HD.

This workflow can include filtering, normalization, log transformation, highly variable gene selection, scaling, and PCA-compatible preparation.

Spatial domain identification

`spartan_spatial_domains`

sp.tl.spartan_spatial_domains(
    adata,
    spatial_coord="grid",
    spatial_neighs=6,
    spatial_rings=2,
    spatial_neighborhood="knn",
    total_pca_comps=50,
    pca_comps_extract=30,
    gene_coord="generic",
    gene_neighs=15,
    alpha=0.80,
    beta1=0.10,
    beta2=0.40,
    resolution=1.0,
    seed=1,
    key_added="spartan_domains",
    copy=False,
)

Runs the complete Spartan spatial domain workflow:

Constructs the spatial graph S.
Constructs the Local Spatial Activation graph L.
Constructs the gene expression connectivity graph G.
Forms the aggregated graph J.
Runs Leiden clustering.
Stores spatial domains in adata.obs[key_added].

The aggregated graph is:

$$J = (\alpha-\beta_1)L + (1-\alpha)G + (\alpha-\beta_2)S$$

where:

L is the Local Spatial Activation graph,
G is the gene expression connectivity graph,
S is the spatial adjacency graph.

Key parameters

Parameter	Description
`spatial_coord`	Coordinate mode used by Squidpy. Common values are `"grid"` for Visium-like data and `"generic"` for single-cell imaging data.
`spatial_neighborhood`	Spatial graph construction method: `"knn"` or `"delaunay"`.
`spatial_neighs`	Number of spatial neighbors for KNN-based spatial graph construction.
`spatial_rings`	Number of spatial rings for grid-based datasets.
`total_pca_comps`	Total number of principal components computed.
`pca_comps_extract`	Number of principal components used for graph construction.
`gene_coord`	Coordinate mode used for gene expression graph construction.
`gene_neighs`	Number of neighbors for the gene expression connectivity graph.
`alpha`	Graph integration parameter controlling the balance between activation/spatial structure and expression connectivity.
`beta1`	Offset controlling the effective LSA graph contribution.
`beta2`	Offset controlling the effective spatial graph contribution.
`resolution`	Leiden resolution parameter controlling clustering granularity.
`seed`	Random seed for reproducibility.
`key_added`	Column name used to store Spartan domain labels in `adata.obs`.
`copy`	If `True`, returns a modified copy of `adata`; otherwise updates `adata` in place.

By default, beta1 + beta2 = 0.5.

Recommended parameter settings

The table below provides suggested starting settings for major spatial transcriptomics technologies. These values are intended as practical defaults for new users and can be adjusted depending on tissue complexity, dataset size, spatial resolution, and desired spatial domain granularity.

Technology	`spatial_coord`	`spatial_neighborhood`	`spatial_neighs`	`spatial_rings`	`gene_coord`	`gene_neighs`	Suggested `alpha`	Suggested `beta1`	Suggested `beta2`	Notes
Visium HD	`grid`	`knn`	4–6	2	`generic`	15	0.70–0.80	0.10–0.26	0.24–0.40	Use higher `alpha` values for large high-resolution datasets. For microstructure detection, explore lower `beta1` values to increase the effective LSA contribution.
MERFISH	`generic`	`knn`	10–12	NA	`generic`	15	0.69–0.82	0.10–0.26	0.24–0.40	Recommended for single-cell imaging datasets. Use higher `alpha` and lower `beta1` values when local spatial activation is expected to be highly informative.
MERFISH	`generic`	`delaunay`	NA	NA	`generic`	15	0.75–0.85	0.10–0.26	0.24–0.40	Delaunay graph construction can be useful for imaging-based single-cell datasets with irregular spatial layouts.
Stereo-seq	`grid`	`knn`	4	1	`generic`	15	0.50–0.60	0.26	0.24	Sequencing-based datasets often stabilize at lower `alpha` values, reflecting a more balanced contribution of LSA, spatial adjacency, and gene-expression connectivity.
Visium SD	`grid`	`knn`	6	2	`generic`	15	0.55–0.75	0.26	0.24	Standard Visium spot-level datasets generally benefit from moderate `alpha` values.

Parameter notes

The recommended settings above are starting points, not strict rules. Users can adjust parameters depending on the biological question and dataset resolution.

The following parameters are especially useful to tune:

Parameter	Default	Description
`gene_neighs`	15	Number of neighbors used to construct the gene-expression connectivity graph.
`total_pca_comps`	50	Total number of principal components computed during PCA.
`pca_comps_extract`	30	Number of principal components used for graph construction.

A good default setting for most datasets is:

gene_neighs = 15
total_pca_comps = 50
pca_comps_extract = 30

Graph construction only

`spartan_build_graphs`

sp.tl.spartan_build_graphs(
    adata,
    spatial_coord="grid",
    spatial_neighs=6,
    spatial_rings=2,
    spatial_neighborhood="knn",
    total_pca_comps=50,
    pca_comps_extract=30,
    gene_coord="generic",
    gene_neighs=15,
    seed=1,
    copy=False,
)

Builds and stores Spartan graph components without running Leiden clustering.

This is useful for:

inspecting graph structure,
running custom parameter sweeps,
alpha-selection workflows,
reusing precomputed graphs,
separating graph construction from clustering.

The function stores:

adata.obsp["spartan_spatial_graph"]
adata.obsp["spartan_spatial_weights"]
adata.obsp["spartan_lsa_graph"]
adata.obsp["spartan_gene_graph"]

Alpha-selection and operating-regime analysis

Spartan includes utilities for evaluating dataset-level graph integration regimes across alpha and resolution values.

These workflows are intended for benchmarking-oriented analysis when ground-truth annotations are available. Metrics such as NMI, homogeneity, and completeness are used to characterize stable operating regimes.

Importantly, Spartan spatial domain detection itself remains unsupervised: graph construction, graph integration, Leiden clustering, and nLSAS-based pruning do not use ground-truth labels.

`initiate_alpha_selection`

summary_df, results_df = sp.tl.initiate_alpha_selection(
    adata,
    lower_alpha=0.50,
    upper_alpha=0.90,
    step_alpha=0.01,
    lower_resolution=0.50,
    upper_resolution=2.00,
    step_resolution=0.05,
    ground_truth="ground_truth",
    config="lsg",
    seed=1,
)

Performs alpha and resolution analysis for a given graph-integration configuration.

The workflow can be used to evaluate stable dataset-level alpha regimes rather than isolated sample-specific optima.

`consensus_alpha`

alpha_star = sp.tl.consensus_alpha(summary_df)

Computes a dataset-level consensus alpha from alpha-selection summaries.

nLSAS-based pruning

Spartan supports nLSAS-based filtering of candidate configurations after graph construction and clustering.

nLSAS is used as an unsupervised stability/coherence criterion to reduce the configuration space before downstream benchmarking and visualization.

Ground-truth labels are not used in nLSAS pruning.

Spatially variable gene discovery

`spartan_svg`

sp.tl.spartan_svg(
    adata,
    lsa_graph=adata.obsp["spartan_lsa_graph"],
    layer="log1pX",
    n_permutations=1000,
    n_cores=8,
    alpha_svg=0.05,
    chunk_size=200,
    seed=1,
    key_added="spartan_svg",
    copy=False,
)

Runs Spartan spatially variable gene discovery using the Spatial Activation Quotient (SAQ).

SAQ measures how strongly each gene aligns with the Local Spatial Activation graph.

The function stores the following columns in adata.var:

adata.var["spartan_saq"]
adata.var["spartan_saq_pval"]
adata.var["spartan_saq_fdr"]
adata.var["spartan_svg"]
adata.var["spartan_saq_rank"]

Key parameters for `sp.tl.spartan_svg`

sp.tl.spartan_svg performs SAQ-based spatially variable gene discovery using the Local Spatial Activation graph. For each gene, Spartan computes a Spatial Activation Quotient (SAQ) score that measures how strongly the gene expression pattern aligns with the LSA graph. Statistical significance is estimated using permutation-based null distributions followed by Benjamini–Hochberg FDR correction.

Parameter	Default	Description
`adata`	required	Input `AnnData` object containing expression values and gene metadata.
`lsa_graph`	required	Local Spatial Activation graph, usually `adata.obsp["spartan_lsa_graph"]`, used to score spatial activation of each gene.
`layer`	`"log1pX"`	Expression layer used for SAQ scoring. If the layer is unavailable and `use_X_if_missing=True`, Spartan falls back to `adata.X`.
`n_permutations`	`1000`	Total number of permutations used to estimate the null distribution of SAQ scores. Higher values provide more stable p-values but increase runtime.
`n_cores`	`8`	Number of CPU cores used for parallel permutation testing.
`use_X_if_missing`	`True`	If `True`, uses `adata.X` when the specified `layer` is not found in `adata.layers`. If `False`, an error is raised when the layer is missing.
`alpha_svg`	`0.05`	FDR threshold used to call significant spatially variable genes. Genes with `spartan_saq_fdr < alpha_svg` are marked as SVGs.
`chunk_size`	`200`	Number of genes processed per chunk. Smaller chunks reduce memory usage; larger chunks may improve speed on machines with sufficient memory.
`seed`	`1`	Random seed used for reproducible permutation testing.
`key_added`	`"spartan_svg"`	Column name added to `adata.var` containing Boolean SVG calls.
`copy`	`False`	If `True`, returns a modified copy of `adata`; otherwise updates `adata` in place.
`dtype`	`np.float32`	Dense array type used during chunked computation. `float32` reduces memory usage and is usually sufficient.
`prefer_backend`	`"threads"`	Joblib backend used for parallel permutation testing. Use `"threads"` for lower overhead; use `"processes"` if thread contention becomes an issue.
`two_stage`	`True`	Enables two-stage permutation testing. A smaller first-stage permutation run is applied to all genes, followed by refinement of the top-ranked candidates.
`n_permutations_stage1`	`100`	Number of first-stage permutations applied to all genes when `two_stage=True`.
`top_k_refine`	`3000`	Number of top candidate genes refined using the remaining permutations in stage 2. Smaller values improve speed; larger values refine more candidates.

Recommended usage

For most datasets, the default two-stage mode provides a good balance between speed and statistical stability:

sp.tl.spartan_svg(
    adata,
    lsa_graph=adata.obsp["spartan_lsa_graph"],
    layer="log1pX",
    n_permutations=1000,
    n_cores=8,
    alpha_svg=0.05,
    key_added="spartan_svg",
)

Plotting API

Spartan provides lightweight plotting helpers in spartan.pl.

Spatial domains

sp.pl.spatial_domains(
    adata,
    color="spartan_domains",
)

SVG table

sp.pl.svg_table(
    adata,
    n=20,
)

For publication-quality plots, Spartan outputs can also be visualized using Scanpy, Squidpy, Matplotlib, or SpatialData plotting utilities.

Tutorial notebooks

Reviewer-oriented tutorial notebooks are available in the tutorials/ directory.

Notebook	Description
`ImagingBasedSpartan.ipynb`	MERFISH imaging-based workflow tutorial
`SequencingBasedSpartan.ipynb`	Stereo-seq sequencing-based workflow tutorial
`VisiumHDAnalysis Using Spartan.ipynb`	High-resolution Visium HD analysis and main figure reproduction
`SpartanSVGDiscovery.ipynb`	Spartan's SVG discovery results and main figure reproduction
`Generate_Benchmarking_and_Ablation_Study_Figures tutorial.ipynb`	Detailed notebook to generate ablation and benchmarking studies panel figures in Figure 3 of the paper

These notebooks demonstrate:

imaging-based spatial domain analysis,
sequencing-based spatial domain analysis,
high-resolution Visium HD biological interpretation,
dataset-level alpha operating-regime analysis,
nLSAS-based configuration filtering,
reproduction of key manuscript analyses and figure panels.
Spartan's SVG discovery utility across diverse spatially resolved transcriptomics technologies.

Full documentation is available on Read the Docs. The README provides the complete quickstart and reviewer-facing overview, while Read the Docs provides structured navigation for installation, API reference, spatial domain identification, SVG discovery, and reproducibility.

Reproducibility

Spartan provides two environments:

General user environment

conda env create -f envs/environment.core.yml
conda activate spartan-core
pip install -e .

Paper reproduction environment

conda env create -f envs/environment.paper.lock.yml
conda activate spartan-paperS
pip install -e .

The paper-lock environment is intended for reproducing manuscript analyses and tutorial notebooks.

Data availability for tutorials and reproducibility

The primary data accession information has been submitted to ArrayExpress and is reported in the manuscript. The SRT samples and processed analysis files used by the tutorial notebooks, including .zarr, .h5ad, and .csv files, are provided for peer review as a password-protected archive.

The archive password is the manuscript number. This temporary archive is provided for peer review to help reviewers reproduce the tutorial notebooks and manuscript figure-generation workflows without needing to regenerate intermediate files or download individual SRT samples from separate public repositories. All files required for reproducing the tutorial workflows are provided in one place.

After publication, the processed reproducibility files will be linked through the permanent public accession or repository associated with the manuscript.

Tested environment

The core package has been tested with:

numpy 2.2.6
scipy 1.15.2
anndata 0.11.4
scanpy 1.11.4
squidpy 1.6.5
igraph 0.11.8
leidenalg 0.10.2
joblib 1.5.1
statsmodels 0.14.5
matplotlib 3.10.5
spatialdata 0.4.0

Citation

If you use Spartan, please cite:

@article {Faiz2026.02.18.706570,
	author = {Faiz, Mohammad Faiz Iqbal and Jokl, Elliot and Jennings, Rachel and Piper Hanley, Karen and Sharrocks, Andrew and Iqbal, Mudassar and Baker, Syed Murtuza},
	title = {Spartan: activation-aware framework for spatial domain and variable gene discovery},
	elocation-id = {2026.02.18.706570},
	year = {2026},
	doi = {10.64898/2026.02.18.706570},
	publisher = {Cold Spring Harbor Laboratory},
	abstract = {Spatial transcriptomics is rapidly advancing toward single-cell-level resolution, revealing complex tissue architectures organized across continuous anatomical gradients. However, accurate identification of spatial domains remains a central computational challenge, as many existing clustering approaches blur anatomical boundaries, merge transitional zones, or fail to resolve localized microstructures. Here we introduce Spartan, an activation-aware multiplex graph framework for high-resolution domain discovery. Spartan integrates spatial topology and Local Spatial Activation (LSA), a neighborhood deviation signal that captures localized transcriptional heterogeneity often attenuated by similarity-based clustering. By jointly modeling cohesion within domains and localized activation structure, Spartan recovers anatomically aligned partitions across spatially resolved transcriptomics technologies including Visium HD, MERFISH, Stereo-seq, and STARmap. We further demonstrate its utility in a high-resolution Visium HD section of developing human esophagus and stomach, where activation-aware graph integration enables precise delineation of complex transitional regions such as the gastroesophageal junction and supports stable multi-scale domain recovery without fragile hyperparameter tuning. Beyond domain identification, Spartan leverages activation-aware structure to detect spatially variable genes associated with localized tissue remodeling. Spartan scales near-linearly with dataset size, providing a robust and interpretable framework for spatial systems-level analysis.Competing Interest StatementThe authors have declared no competing interest.BBSRC DTP},
	URL = {https://www.biorxiv.org/content/early/2026/04/30/2026.02.18.706570},
	eprint = {https://www.biorxiv.org/content/early/2026/04/30/2026.02.18.706570.full.pdf},
	journal = {bioRxiv}
}

License

This project is released under the license specified in the repository.

Contact

For questions, issues, or contributions, please open an issue on GitHub:

https://github.com/MohammadFaizIqbalFaiz/spartan-st

Name		Name	Last commit message	Last commit date
Latest commit History 145 Commits
assets		assets
docs		docs
envs		envs
src/spartan		src/spartan
tutorials		tutorials
.gitattributes		.gitattributes
.gitignore		.gitignore
.readthedocs.yaml		.readthedocs.yaml
CITATION.cff		CITATION.cff
LICENSE		LICENSE
README.md		README.md
pyproject.toml		pyproject.toml
requirements.txt		requirements.txt

Folders and files

Latest commit

History

Repository files navigation

Spartan

Overview

Installation

General package environment

Paper reproducibility environment

Quickstart tutorial

1. Import packages and load a MERFISH sample

2. Add a ground-truth column if needed

3. Preprocess the imaging-based dataset

4. Run Spartan spatial domain identification

5. Run SAQ-based spatially variable gene discovery

6. Visualize cell classes and Spartan spatial domains

7. Visualize representative Spartan SVGs

8. Main objects created by the quickstart

Complete quickstart code

Main outputs

Spatial domains

Graphs stored in adata.obsp

Spatially variable gene outputs

Core user-facing API

Preprocessing

pre_process_imaging

pre_process_sequencing

Spatial domain identification

spartan_spatial_domains

Key parameters

Recommended parameter settings

Parameter notes

Graph construction only

spartan_build_graphs

Alpha-selection and operating-regime analysis

initiate_alpha_selection

consensus_alpha

nLSAS-based pruning

Spatially variable gene discovery

spartan_svg

Key parameters for sp.tl.spartan_svg

Recommended usage

Plotting API

Spatial domains

SVG table

Tutorial notebooks

Reproducibility

General user environment

Paper reproduction environment

Data availability for tutorials and reproducibility

Tested environment

Citation

License

Contact

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages

Graphs stored in `adata.obsp`

`pre_process_imaging`

`pre_process_sequencing`

`spartan_spatial_domains`

`spartan_build_graphs`

`initiate_alpha_selection`

`consensus_alpha`

`spartan_svg`

Key parameters for `sp.tl.spartan_svg`