Chromium PBMC Landscape + Clustergram + Enrich¶

This tutorial starts with a public 10x Genomics Chromium count matrix and follows the Scanpy-to-Celldega workflow:

1. preprocess the cells, compute a UMAP, and find Leiden clusters with Scanpy,
2. organize the clusters with a SetCollection, retaining cell membership, expression signatures, fraction expressing, and ranked marker genes,
3. explore marker-ranked Clustergram views linked to a Landscape UMAP and Enrich,
4. annotate clusters and genes interactively, and
5. cluster individual cells in a single-cell Clustergram linked to Enrich.

Data: PBMCs from a Healthy Donor: Whole Transcriptome Analysis, provided by 10x Genomics under CC BY 4.0. The filtered matrix contains 6,574 cells. It is downloaded once and cached locally.

Setup¶

Use Python 3.11 or 3.12 and Celldega 0.26.0 or later. Install the notebook dependencies in the environment used by your Jupyter kernel:

%pip install "celldega[scanpy]>=0.26.0" igraph ipykernel

The Landscape reads hosted expression files for this same sample, and Enrich queries the public Enrichr service. Both need an internet connection. The saved interactive widgets on this documentation page work without a running Python kernel.

In [1]:

from importlib.metadata import version
from pathlib import Path
import shutil
from urllib.request import Request, urlopen
import warnings

import numpy as np
import pandas as pd
import scanpy as sc

import celldega as dega


print(f"celldega {version('celldega')}")
print(f"scanpy {version('scanpy')}")
sc.settings.verbosity = 1

from importlib.metadata import version from pathlib import Path import shutil from urllib.request import Request, urlopen import warnings import numpy as np import pandas as pd import scanpy as sc import celldega as dega print(f"celldega {version('celldega')}") print(f"scanpy {version('scanpy')}") sc.settings.verbosity = 1

celldega 0.26.0
scanpy 1.10.4

Download and read the 10x count matrix¶

This is the Cell Ranger 4.0.0 filtered gene-expression matrix from the dataset above. The HDF5 file is about 15 MB. Downloads use a temporary file so an interrupted transfer is not reused as a complete matrix.

In [2]:

DATASET_URL = (
    "https://cf.10xgenomics.com/samples/cell-vdj/4.0.0/Parent_SC5_PBMC/"
    "Parent_SC5_PBMC_filtered_feature_bc_matrix.h5"
)

# Reuse the repository's notebook data directory when running from a checkout;
# otherwise cache the download beside the notebook.
working_dir = Path.cwd().resolve()
project_root = next(
    (
        directory
        for directory in (working_dir, *working_dir.parents)
        if (directory / "pyproject.toml").is_file() and (directory / "notebooks").is_dir()
    ),
    None,
)
data_dir = (project_root / "notebooks" if project_root else working_dir) / "data"
data_dir = data_dir / "chromium_data" / "chromium_pbmc_healthy-donor"
data_dir.mkdir(parents=True, exist_ok=True)
counts_path = data_dir / "Parent_SC5_PBMC_filtered_feature_bc_matrix.h5"

if not counts_path.is_file():
    partial = counts_path.with_suffix(".h5.part")
    request = Request(DATASET_URL, headers={"User-Agent": "celldega-example-notebook"})
    print("Downloading the 10x PBMC matrix...")
    try:
        with urlopen(request, timeout=300) as response, partial.open("wb") as output:
            shutil.copyfileobj(response, output, length=1024 * 1024)
        partial.replace(counts_path)
    except Exception:
        partial.unlink(missing_ok=True)
        raise

with warnings.catch_warnings():
    # The 10x matrix repeats some gene symbols; make them unique below.
    warnings.filterwarnings("ignore", message="Variable names are not unique")
    adata = sc.read_10x_h5(counts_path, gex_only=True)
adata.var_names_make_unique()
adata

DATASET_URL = ( "https://cf.10xgenomics.com/samples/cell-vdj/4.0.0/Parent_SC5_PBMC/" "Parent_SC5_PBMC_filtered_feature_bc_matrix.h5" ) # Reuse the repository's notebook data directory when running from a checkout; # otherwise cache the download beside the notebook. working_dir = Path.cwd().resolve() project_root = next( ( directory for directory in (working_dir, *working_dir.parents) if (directory / "pyproject.toml").is_file() and (directory / "notebooks").is_dir() ), None, ) data_dir = (project_root / "notebooks" if project_root else working_dir) / "data" data_dir = data_dir / "chromium_data" / "chromium_pbmc_healthy-donor" data_dir.mkdir(parents=True, exist_ok=True) counts_path = data_dir / "Parent_SC5_PBMC_filtered_feature_bc_matrix.h5" if not counts_path.is_file(): partial = counts_path.with_suffix(".h5.part") request = Request(DATASET_URL, headers={"User-Agent": "celldega-example-notebook"}) print("Downloading the 10x PBMC matrix...") try: with urlopen(request, timeout=300) as response, partial.open("wb") as output: shutil.copyfileobj(response, output, length=1024 * 1024) partial.replace(counts_path) except Exception: partial.unlink(missing_ok=True) raise with warnings.catch_warnings(): # The 10x matrix repeats some gene symbols; make them unique below. warnings.filterwarnings("ignore", message="Variable names are not unique") adata = sc.read_10x_h5(counts_path, gex_only=True) adata.var_names_make_unique() adata

Out[2]:

AnnData object with n_obs × n_vars = 6574 × 36601
    var: 'gene_ids', 'feature_types', 'genome', 'pattern', 'read', 'sequence'

Quality control and Scanpy clustering¶

Calculate mitochondrial, ribosomal, and hemoglobin QC metrics; retain cells with at least 100 detected genes and genes detected in at least three cells. These minimal filters keep the original tutorial's cell selection. Inspect QC metrics and choose appropriate thresholds when applying this workflow to your own data.

Keep raw counts in adata.layers["counts"]. Normalize and log-transform X for PCA, neighbor finding, UMAP, and marker testing. Highly variable genes drive PCA, but all retained genes remain available for signatures and markers.

In [3]:

adata.var["mt"] = adata.var_names.str.startswith("MT-")
adata.var["ribo"] = adata.var_names.str.startswith(("RPS", "RPL"))
adata.var["hb"] = adata.var_names.str.contains(r"^HB(?!P)")
sc.pp.calculate_qc_metrics(adata, qc_vars=["mt", "ribo", "hb"], inplace=True)

sc.pp.filter_cells(adata, min_genes=100)
sc.pp.filter_genes(adata, min_cells=3)
adata.layers["counts"] = adata.X.copy()

print(f"Retained {adata.n_obs:,} cells and {adata.n_vars:,} genes")
adata.obs[["n_genes_by_counts", "total_counts", "pct_counts_mt"]].describe().round(2)

adata.var["mt"] = adata.var_names.str.startswith("MT-") adata.var["ribo"] = adata.var_names.str.startswith(("RPS", "RPL")) adata.var["hb"] = adata.var_names.str.contains(r"^HB(?!P)") sc.pp.calculate_qc_metrics(adata, qc_vars=["mt", "ribo", "hb"], inplace=True) sc.pp.filter_cells(adata, min_genes=100) sc.pp.filter_genes(adata, min_cells=3) adata.layers["counts"] = adata.X.copy() print(f"Retained {adata.n_obs:,} cells and {adata.n_vars:,} genes") adata.obs[["n_genes_by_counts", "total_counts", "pct_counts_mt"]].describe().round(2)

Retained 6,565 cells and 18,314 genes

Out[3]:

	n_genes_by_counts	total_counts	pct_counts_mt
count	6565.00	6565.00	6565.00
mean	1555.94	5014.09	6.27
std	624.40	3251.18	4.78
min	102.00	506.00	0.00
25%	1229.00	3353.00	4.33
50%	1473.00	4470.00	5.31
75%	1744.00	5728.00	6.57
max	6042.00	47031.00	82.33

In [4]:

sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.log1p(adata)
sc.pp.highly_variable_genes(adata, n_top_genes=2000)
sc.tl.pca(adata, mask_var="highly_variable", svd_solver="arpack", random_state=0)
sc.pp.neighbors(adata, n_neighbors=15, n_pcs=40, random_state=0)
sc.tl.umap(adata, random_state=0)
sc.tl.leiden(
    adata,
    flavor="igraph",
    directed=False,
    n_iterations=2,
    resolution=1.0,
    random_state=0,
    key_added="leiden",
)

sc.pp.normalize_total(adata, target_sum=1e4) sc.pp.log1p(adata) sc.pp.highly_variable_genes(adata, n_top_genes=2000) sc.tl.pca(adata, mask_var="highly_variable", svd_solver="arpack", random_state=0) sc.pp.neighbors(adata, n_neighbors=15, n_pcs=40, random_state=0) sc.tl.umap(adata, random_state=0) sc.tl.leiden( adata, flavor="igraph", directed=False, n_iterations=2, resolution=1.0, random_state=0, key_added="leiden", )

Plotting the UMAP stores Scanpy's cluster palette in adata.uns["leiden_colors"]. The Landscape, SetCollection, and Clustergram reuse this palette, so each cluster has the same color throughout the tutorial.

In [5]:

sc.pl.umap(adata, color="leiden", legend_loc="on data")

sc.pl.umap(adata, color="leiden", legend_loc="on data")

SetCollection: signatures, fraction expressing, and marker genes¶

Each Leiden cluster becomes one set. The sparse membership modality records which cells belong to it, and setc.obs records the cluster's cell count and color.

In [6]:

setc = dega.SetCollection(adata, set_col="leiden", name="leiden")
setc.obs[["leiden", "n_cells", "color"]]

setc = dega.SetCollection(adata, set_col="leiden", name="leiden") setc.obs[["leiden", "n_cells", "color"]]

Out[6]:

	leiden	n_cells	color
leiden
0	0	716	#1f77b4
2	2	1330	#279e68
3	3	285	#d62728
4	4	593	#aa40fc
5	5	712	#8c564b
6	6	102	#e377c2
1	1	1308	#ff7f0e
7	7	716	#b5bd61
8	8	182	#17becf
9	9	64	#aec7e8
10	10	501	#ffbb78
11	11	49	#98df8a
12	12	7	#ff9896

calc_signature attaches an expression modality with sets as observations and genes as variables:

• X: summed raw counts per cluster, normalized to counts per million and log1p-transformed (log1p_cpm pseudobulk signatures).
• layers["fraction_expressing"]: the fraction of cells in each cluster with nonzero raw counts for each gene, used for dot size.
• uns["rank_genes_groups"]: Scanpy marker rankings computed on the cell-level, log-normalized adata.X, using a Wilcoxon test against the remaining cells.
• var["leiden_marker"]: each gene's highest-scoring cluster, used to color rows.

layer="counts" selects the aggregation and fraction-expressing source; rank_genes_groups_layer="X" explicitly selects the normalized source for marker testing. The method prints these sources. Marker rankings are retained for every gene; they are exploratory cluster markers, not a comparison between donors.

In [7]:

setc.calc_signature(
    adata,
    modality_name="expression",
    layer="counts",
    aggregate="sum",
    normalization="log1p_cpm",
    fraction_expressing_layer="fraction_expressing",
    rank_genes_groups=True,
    rank_genes_groups_layer="X",
    rank_genes_groups_kwargs={"method": "wilcoxon", "tie_correct": True},
)
setc.mod["expression"]

setc.calc_signature( adata, modality_name="expression", layer="counts", aggregate="sum", normalization="log1p_cpm", fraction_expressing_layer="fraction_expressing", rank_genes_groups=True, rank_genes_groups_layer="X", rank_genes_groups_kwargs={"method": "wilcoxon", "tie_correct": True}, ) setc.mod["expression"]

calc_signature('expression'): sum of adata.layers['counts'] (normalization='log1p_cpm')
  layers['fraction_expressing']: fraction of cells with adata.layers['counts'] > 0.0
  uns['rank_genes_groups']: wilcoxon on adata.X, grouped by 'leiden'

Out[7]:

AnnData object with n_obs × n_vars = 13 × 18314
    obs: 'leiden', 'n_cells', 'set_source', 'color'
    var: 'gene_ids', 'feature_types', 'genome', 'pattern', 'read', 'sequence', 'mt', 'ribo', 'hb', 'n_cells_by_counts', 'mean_counts', 'log1p_mean_counts', 'pct_dropout_by_counts', 'total_counts', 'log1p_total_counts', 'n_cells', 'highly_variable', 'means', 'dispersions', 'dispersions_norm', 'gene', 'leiden_marker', 'entity_type'
    uns: 'feature_type', 'aggregate', 'normalization', 'layer', 'expr_threshold', 'fraction_expressing_layer', 'rank_genes_groups', 'axis_entities'
    layers: 'fraction_expressing'

In [8]:

# Celldega stores the Scanpy results in a tidy, serializable column-wise form.
markers = pd.DataFrame(setc.mod["expression"].uns["rank_genes_groups"])
markers.loc[markers["rank"] < 3, ["group", "names", "scores", "logfoldchanges", "pvals_adj"]]

# Celldega stores the Scanpy results in a tidy, serializable column-wise form. markers = pd.DataFrame(setc.mod["expression"].uns["rank_genes_groups"]) markers.loc[markers["rank"] < 3, ["group", "names", "scores", "logfoldchanges", "pvals_adj"]]

Out[8]:

	group	names	scores	logfoldchanges	pvals_adj
0	0	GZMK	54.846451	5.621858	0.000000e+00
1	0	CCL5	46.330376	4.899702	0.000000e+00
2	0	DUSP2	34.398769	3.292818	1.605802e-255
18314	1	RPL32	38.829815	0.920654	0.000000e+00
18315	1	RPS3A	38.652439	0.908228	0.000000e+00
18316	1	RPL30	38.433609	0.830259	0.000000e+00
36628	10	MS4A1	72.087128	8.787992	0.000000e+00
36629	10	LINC00926	70.512154	8.741061	0.000000e+00
36630	10	CD79A	68.571426	8.514773	0.000000e+00
54942	11	LRRC26	76.926704	13.226185	0.000000e+00
54943	11	AC097375.1	71.286835	30.928379	0.000000e+00
54944	11	PTCRA	69.563103	11.408402	0.000000e+00
73256	12	GP9	75.791641	18.838745	0.000000e+00
73257	12	ITGB3	75.788338	15.717124	0.000000e+00
73258	12	CMTM5	75.002846	33.501492	0.000000e+00
91570	2	IL7R	33.449158	2.281880	4.848532e-241
91571	2	LTB	32.986423	1.951516	1.164534e-234
91572	2	ITGB1	32.545738	2.358950	1.468537e-228
109884	3	MT-CO3	27.147522	1.954384	4.960209e-158
109885	3	MT-ATP6	26.908867	1.980738	1.583718e-155
109886	3	MT-CYB	26.507257	2.054489	1.626720e-151
128198	4	LINC02446	57.216511	5.883980	0.000000e+00
128199	4	CD8B	57.018902	5.520812	0.000000e+00
128200	4	CD8A	44.567101	4.007354	0.000000e+00
146512	5	GZMB	69.860069	7.541908	0.000000e+00
146513	5	GNLY	67.074066	8.906074	0.000000e+00
146514	5	FGFBP2	66.153435	8.042933	0.000000e+00
164826	6	FCER1A	58.387558	8.915568	0.000000e+00
164827	6	ENHO	53.217014	8.127621	0.000000e+00
164828	6	CLEC10A	49.691978	7.676640	0.000000e+00
183140	7	S100A12	69.300377	9.487892	0.000000e+00
183141	7	S100A8	69.269867	9.781773	0.000000e+00
183142	7	VCAN	68.801338	7.647643	0.000000e+00
201454	8	CDKN1C	62.900661	8.043122	0.000000e+00
201455	8	HES4	52.532467	6.671907	0.000000e+00
201456	8	AC104809.2	50.822643	7.715016	0.000000e+00
219768	9	SPTSSB	39.295662	7.293156	0.000000e+00
219769	9	KLRC1	31.671799	6.298415	3.479976e-216
219770	9	XCL1	25.369282	5.465434	3.361628e-138

Clustergram with marker-ranked views¶

The Matrix has genes as rows and Leiden clusters as columns. Its colors show each gene's z-scored pseudobulk expression across clusters, while dot sizes show the fraction of cells expressing it. Z-scoring changes only the Matrix; the SetCollection retains the original signatures and fractions.

cluster(view="rank_genes_groups", levels=...) precomputes a view for each requested number of top markers per cluster. Genes shared by multiple clusters appear once, and both axes are hierarchically clustered again for each view. The DIM slider switches between these views and the full gene matrix. rank_dim=3 starts at the top three markers per cluster.

In [9]:

marker_levels = [1, 2, 3, 5, 10, 25, 50, 75]

mat = dega.clust.Matrix(
    collection=setc,
    color_by="expression",
    size_by_layer="fraction_expressing",
    col_attr=["leiden"],
    row_attr=["leiden_marker"],
)
mat.norm("row", by="zscore")
mat.cluster(view="rank_genes_groups", levels=marker_levels)

for view in mat.views:
    print(f"{view['level']:>3} markers/cluster -> {view['n_rows']} genes")
print(f"all -> {mat.data.shape[0]:,} genes")

cgm = dega.viz.Clustergram(
    matrix=mat,
    width=450,
    height=500,
    viz_mode="dotplot",
    manual_col_cat="cell_type",
    manual_row_cat="info",
    rank_dim=3,
)

marker_levels = [1, 2, 3, 5, 10, 25, 50, 75] mat = dega.clust.Matrix( collection=setc, color_by="expression", size_by_layer="fraction_expressing", col_attr=["leiden"], row_attr=["leiden_marker"], ) mat.norm("row", by="zscore") mat.cluster(view="rank_genes_groups", levels=marker_levels) for view in mat.views: print(f"{view['level']:>3} markers/cluster -> {view['n_rows']} genes") print(f"all -> {mat.data.shape[0]:,} genes") cgm = dega.viz.Clustergram( matrix=mat, width=450, height=500, viz_mode="dotplot", manual_col_cat="cell_type", manual_row_cat="info", rank_dim=3, )

  1 markers/cluster -> 13 genes
  2 markers/cluster -> 26 genes
  3 markers/cluster -> 39 genes
  5 markers/cluster -> 65 genes
 10 markers/cluster -> 124 genes
 25 markers/cluster -> 298 genes
 50 markers/cluster -> 561 genes
 75 markers/cluster -> 805 genes
all -> 18,314 genes

Landscape: view the same cells in UMAP space¶

Chromium cells have no tissue coordinates, so the Landscape displays the UMAP computed above. It takes cell labels, colors, and the embedding from the same adata used to build the SetCollection. The hosted LandscapeFiles supply per-gene counts for this PBMC sample; they do not supply the newly computed clusters.

The signature records that columns are cells grouped by leiden, allowing a Clustergram column click to select the matching cells in the Landscape.

In [10]:

base_url = "https://raw.githubusercontent.com/broadinstitute/chromium_pbmc_healthy-donor/main"

meta_cluster = setc.obs[["color"]].copy()
meta_cluster["count"] = setc.obs["n_cells"]
landscape = dega.viz.Landscape(
    technology="Chromium",
    base_url=base_url,
    adata=adata,
    meta_cluster=meta_cluster,
    cell_attr=["leiden"],
    cluster_attr="leiden",
    # Chromium UMAPs do not need a tissue-image coordinate transform.
    transform=np.eye(3),
    height=600,
    ini_zoom=-4.5,
    ini_x=4000,
    ini_y=5000,
)

base_url = "https://raw.githubusercontent.com/broadinstitute/chromium_pbmc_healthy-donor/main" meta_cluster = setc.obs[["color"]].copy() meta_cluster["count"] = setc.obs["n_cells"] landscape = dega.viz.Landscape( technology="Chromium", base_url=base_url, adata=adata, meta_cluster=meta_cluster, cell_attr=["leiden"], cluster_attr="leiden", # Chromium UMAPs do not need a tissue-image coordinate transform. transform=np.eye(3), height=600, ini_zoom=-4.5, ini_x=4000, ini_y=5000, )

Linked Landscape, Clustergram, and Enrich¶

• Move the DIM slider to change the number of markers per cluster, or select all to show every retained gene.
• Click a cluster column label to highlight its cells in the UMAP and send its highest-valued genes in the current view to Enrich. This selection uses the displayed z-scores; the DIM views use the Scanpy marker rankings.
• Click a gene row label to show its expression in the UMAP without replacing the current enrichment gene set.
• Select a gene dendrogram group or crop rows to send that gene set to Enrich.
• Use Enrich's library selector to explore cell-type markers or biological processes. Select a term to highlight its genes in the Clustergram; click a gene in Enrich to focus it in both visualizations.
• Use the ROW / COL category controls and annotate selected genes or clusters with the manual info and cell_type categories.

Enrich starts with an empty gene list. Selections submit gene symbols to the public Enrichr API. The widgets link in the browser, including in this saved documentation page.

In [11]:

linked_widgets = dega.viz.spatial_clustergram(
    landscape,
    cgm,
    enrich=True,
    enrich_kwargs={"inst_lib": "CellMarker_2024"},
)

linked_widgets = dega.viz.spatial_clustergram( landscape, cgm, enrich=True, enrich_kwargs={"inst_lib": "CellMarker_2024"}, )

In [12]:

linked_widgets

linked_widgets

Out[12]:

Single-cell Clustergram and Enrich¶

The views above summarize each cluster as a pseudobulk signature. To see the cell-to-cell variation within clusters, build a Matrix directly from adata, with genes as rows and individual cells as columns, colored by their Leiden cluster. The values come from the log-normalized adata.X.

Mitochondrial and ribosomal genes are excluded. The 200 most variable of the 10,000 highest-expressed remaining genes are z-scored across cells, and both axes are hierarchically clustered.

In [13]:

# Exclude mitochondrial and ribosomal genes.
non_mt_ribo = ~adata.var_names.str.upper().str.startswith(("MT-", "RPS", "RPL", "MRPS", "MRPL"))

mat_cells = dega.clust.Matrix(
    adata[:, non_mt_ribo].copy(),
    name="single_cell",
    col_attr=["leiden"],
)
mat_cells.filter(axis="row", by="mean", num=10000)
mat_cells.filter(axis="row", by="var", num=200)
mat_cells.norm(axis="row", by="zscore")
mat_cells.cluster()

cgm_cells = dega.viz.Clustergram(matrix=mat_cells, width=500, height=500)

# Exclude mitochondrial and ribosomal genes. non_mt_ribo = ~adata.var_names.str.upper().str.startswith(("MT-", "RPS", "RPL", "MRPS", "MRPL")) mat_cells = dega.clust.Matrix( adata[:, non_mt_ribo].copy(), name="single_cell", col_attr=["leiden"], ) mat_cells.filter(axis="row", by="mean", num=10000) mat_cells.filter(axis="row", by="var", num=200) mat_cells.norm(axis="row", by="zscore") mat_cells.cluster() cgm_cells = dega.viz.Clustergram(matrix=mat_cells, width=500, height=500)

/Users/feni/Documents/celldega/src/celldega/clust/matrix.py:472: UserWarning: Large matrix (6565 x 18124). Consider filtering.
  self.load_adata(data, col_attr=col_attr, row_attr=row_attr)

Select a gene dendrogram group or crop rows to send those genes to Enrich. Gene groups that separate subsets of cells within a cluster are a starting point for finding cell subtypes or states.

In [14]:

cells_enrich = dega.viz.clustergram_enrich(cgm_cells)

cells_enrich = dega.viz.clustergram_enrich(cgm_cells)

In [15]:

cells_enrich

cells_enrich

Out[15]:

Continue in Python and save the analysis¶

In a live notebook the DIM slider is two-way, and manual annotations sync back to Python:

cgm.rank_dim = 10  # each cluster's top 10 markers
cgm.rank_dim = 0   # all genes

cgm.col_manual_df         # cluster -> cell_type
cgm.row_manual_df         # gene -> info
cgm.col_manual_colors_df  # cell_type -> color

cgm.col_manual_df.to_parquet(data_dir / "pbmc_cluster_annotations.parquet")
cgm.row_manual_df.to_parquet(data_dir / "pbmc_gene_annotations.parquet")

adata.write_h5ad(data_dir / "pbmc_scanpy.h5ad")
setc.write(data_dir / "pbmc_sets.h5mu")

The SetCollection saves membership, signatures, fractions, and marker rankings. Widget annotations are saved separately in the Parquet files above. Reload the collection and construct a Matrix with marker views without rerunning Scanpy:

reloaded = dega.SetCollection.read(data_dir / "pbmc_sets.h5mu")
reloaded_mat = dega.clust.Matrix(
    collection=reloaded,
    color_by="expression",
    size_by_layer="fraction_expressing",
    col_attr=["leiden"],
    row_attr=["leiden_marker"],
)
reloaded_mat.norm("row", by="zscore")
reloaded_mat.cluster(view="rank_genes_groups", levels=marker_levels)

Generate LandscapeFiles for your own Chromium data¶

The hosted files above keep this example interactive on the documentation site. For another dataset, generate matching files from your own adata (including its raw counts layer), then use a local server while running Jupyter on your machine:

path_dega_files = Path("data/landscape_files/my_chromium_sample")
dega.pre.make_chromium_from_anndata(adata, path_dega_files)
port = dega.viz.get_local_server()
base_url = f"http://localhost:{port}/{path_dega_files.as_posix()}"

Use that base_url when constructing the Landscape. A local URL works only while its server is running; to publish an interactive page, host the generated files at a public HTTPS URL and save the widgets with that URL instead.