Vitessce¶

This tutorial has been adopted from the data preparation examples in the Vitessce documention.

We demonstrate how to use Vitessce to create interactive visualizations for data stored in LaminDB Artifacts, using multiple data formats.

# !pip install "vitessce[all]>=3.5.0"
# !pip install "generate-tiff-offsets>=0.1.9"
# !pip install "lamindb[jupyter,aws,bionty]"
!lamin connect laminlabs/lamin-dev  # <-- replace with your instance

from urllib.request import urlretrieve
from pathlib import Path
from anndata import read_h5ad
import vitessce as vit
from vitessce import data_utils as vitdu
from generate_tiff_offsets import get_offsets
import lamindb as ln
import zipfile
import json

# [optional] track the current notebook or script
ln.track("BZhZQ6uIbkWv0001")

Visualize an AnnData object (H5AD format)¶

Here, we download an example H5AD file (approximately 22 MB). This is a dataset from the COVID-19 Cell Atlas.

h5ad_raw_filepath = "./habib17.raw.h5ad"
h5ad_processed_filepath = "./habib17.processed.h5ad"
if not Path(h5ad_raw_filepath).exists():
    urlretrieve(
        "https://covid19.cog.sanger.ac.uk/habib17.processed.h5ad", h5ad_raw_filepath
    )

This AnnData file was not saved recently and read_h5ad will warn us about this. By reading and subsequently re-writing the file using an up-to-date anndata package version, we ensure that the dataset is saved using the latest AnnData H5AD format.

adata = read_h5ad(h5ad_raw_filepath)

/opt/hostedtoolcache/Python/3.12.8/x64/lib/python3.12/site-packages/anndata/compat/__init__.py:358: FutureWarning: Moving element from .uns['neighbors']['distances'] to .obsp['distances'].

This is where adjacency matrices should go now.
  warn(
/opt/hostedtoolcache/Python/3.12.8/x64/lib/python3.12/site-packages/anndata/compat/__init__.py:358: FutureWarning: Moving element from .uns['neighbors']['connectivities'] to .obsp['connectivities'].

This is where adjacency matrices should go now.
  warn(

Optionally, we can modify the AnnData object, for example to subset it to the highly variable genes (which will affect the amount of data that is displayed in the heatmap visualization).

adata = adata[:, adata.var["highly_variable"]].copy()

adata.write_h5ad(h5ad_processed_filepath)

Save the dataset¶

Save the .h5ad file as a LaminDB Artifact.

h5ad_artifact = ln.Artifact(
    h5ad_processed_filepath,
    key="vitessce_examples/habib17.h5ad",
    description="Habib et al., 2017 Nature Methods, h5ad",
    type="dataset",
)
h5ad_artifact.save()

! records with similar descriptions exist! did you mean to load one of them?

	uid	key	description	suffix	type	size	hash	n_objects	n_observations	_hash_type	_accessor	visibility	_key_is_virtual	storage_id	transform_id	version	is_latest	run_id	created_at	created_by_id
id
2805	dhE9utcN0FRdFC3Q0000	vitessce_examples/habib17.adata.zarr	Habib et al., 2017 Nature Methods, zarr	.anndata.zarr	dataset	10601062	Cb1-pcS-n7Ywo2HJ-m7DgA	208	None	md5-d	AnnData	1	True	1	126	None	False	246	2024-12-16 14:07:59.695961+00:00	2
2928	dhE9utcN0FRdFC3Q0001	vitessce_examples/habib17.adata.zarr	Habib et al., 2017 Nature Methods, zarr	.anndata.zarr	dataset	10601062	YPuOwAd5BZ3pbPoGNBH4aw	208	None	md5-d	AnnData	1	True	1	126	None	True	246	2024-12-20 15:02:01.099845+00:00	2

Artifact(uid='DI7oBxKX3oorSAv70000', is_latest=True, key='vitessce_examples/habib17.h5ad', description='Habib et al., 2017 Nature Methods, h5ad', suffix='.h5ad', type='dataset', size=22586523, hash='nBetKKNFs4mjJ04X_f4uaA', _hash_type='md5', _accessor='AnnData', visibility=1, _key_is_virtual=True, storage_id=1, transform_id=126, run_id=246, created_by_id=2, created_at=2024-12-20 15:40:39 UTC)

When using .h5ad files, we construct a Reference Specification which enables interoperability with the Zarr interface.

ref_spec = vitdu.generate_h5ad_ref_spec(h5ad_processed_filepath)

We next need to save the corresponding Reference Specification to a JSON file and upload it to LaminDB as an Artifact.

ref_path = "./habib17.processed.reference.json"
with open(ref_path, "w") as file:
    json.dump(ref_spec, file)
ref_artifact = ln.Artifact(
    ref_path,
    key="vitessce_examples/habib17.reference.json",
    description="Reference JSON for H5AD file, Habib et al., 2017 Nature Methods"
).save()

Save a VitessceConfig object¶

You can create a dashboard for one or several datasets by using Vitessce’s component API.

You can pass LaminDB Artifacts to the AnnDataWrapper class using the adata_artifact and ref_artifact parameters.

vc = vit.VitessceConfig(
    schema_version="1.0.17",
    description=h5ad_artifact.description,
)
dataset = vc.add_dataset(name="Habib 2017").add_object(
    vit.AnnDataWrapper(
        adata_artifact=h5ad_artifact,
        ref_artifact=ref_artifact,
        obs_feature_matrix_path="X",
        obs_embedding_paths=["obsm/X_umap"],
        obs_embedding_names=["UMAP"],
        obs_set_paths=["obs/CellType"],
        obs_set_names=["Cell Type"],
    )
)
obs_sets = vc.add_view(vit.ViewType.OBS_SETS, dataset=dataset)
obs_sets_sizes = vc.add_view(vit.ViewType.OBS_SET_SIZES, dataset=dataset)
scatterplot = vc.add_view(vit.ViewType.SCATTERPLOT, dataset=dataset, mapping="UMAP")
heatmap = vc.add_view(vit.ViewType.HEATMAP, dataset=dataset)
genes = vc.add_view(vit.ViewType.FEATURE_LIST, dataset=dataset)
vc.link_views([scatterplot, heatmap], ["featureValueColormapRange"], [[0.0, 0.1]])
vc.layout(((scatterplot | obs_sets) / heatmap) | (obs_sets_sizes / genes));

Save the VitessceConfig object.

description_habib17_h5ad = "View Habib17 (h5ad) in Vitessce"
h5ad_vc_artifact = ln.integrations.save_vitessce_config(
    vc, description=description_habib17_h5ad,
)

h5ad_vc_artifact.view_lineage()

_images/205960db37160bb3ae2d6d139335794c3e692fb97aa8b0579931581a6d6b4ca0.svg

Note

After running save_vitessce_config, a Vitessce button will appear next to the dataset in the Artifacts page of the web interface.

If your VitessceConfig object references data from multiple artifacts, the Vitessce button will appear next to a Collection that groups these artifacts (on the Collections tab of the Artifacts page). Note that when using an .h5ad-based Artifact, the presence of the corresponding .reference.json file will result in the creation of a Collection.

Visualize an AnnData object (Zarr format)¶

AnnData objects can be saved on-disk to not only .h5ad files, but also to Zarr stores using AnnData’s write_zarr method.

Save the dataset¶

Convert the dataset to .zarr format. Often, Zarr data is stored on-disk using a DirectoryStore. By convention, these directories are named using a .zarr suffix.

zarr_processed_filepath = "./habib17.anndata.zarr"

adata = read_h5ad(h5ad_raw_filepath)

/opt/hostedtoolcache/Python/3.12.8/x64/lib/python3.12/site-packages/anndata/compat/__init__.py:358: FutureWarning: Moving element from .uns['neighbors']['distances'] to .obsp['distances'].

This is where adjacency matrices should go now.
  warn(
/opt/hostedtoolcache/Python/3.12.8/x64/lib/python3.12/site-packages/anndata/compat/__init__.py:358: FutureWarning: Moving element from .uns['neighbors']['connectivities'] to .obsp['connectivities'].

This is where adjacency matrices should go now.
  warn(

Just like in the above section, we may want to modify the AnnData object prior to saving it as a LaminDB Artifact that we will use for visualization.

adata = adata[:, adata.var["highly_variable"]].copy()

Note the usage of .write_zarr below, as opposed to .write_h5ad.

adata.write_zarr(zarr_processed_filepath)

Save the .zarr directory as a LaminDB Artifact.

adata_zarr_artifact = ln.Artifact(
    zarr_processed_filepath,
    key="vitessce_examples/habib17.adata.zarr",
    description="Habib et al., 2017 Nature Methods, zarr",
    type="dataset",
)
adata_zarr_artifact.save()

→ creating new artifact version for key='vitessce_examples/habib17.adata.zarr' (storage: 's3://lamin-eu-central-1/9fm7UN13')

Artifact(uid='dhE9utcN0FRdFC3Q0002', is_latest=True, key='vitessce_examples/habib17.adata.zarr', description='Habib et al., 2017 Nature Methods, zarr', suffix='.anndata.zarr', type='dataset', size=10601062, hash='KWskM8fCvn1QUqq43D5qBQ', n_objects=208, _hash_type='md5-d', _accessor='AnnData', visibility=1, _key_is_virtual=True, storage_id=1, transform_id=126, run_id=246, created_by_id=2, created_at=2024-12-20 15:40:48 UTC)

Save a VitessceConfig object¶

You can create a dashboard for one or several datasets by using Vitessce’s component API. Here, we configure the visualization the same way as above in the .h5ad-based example, with the exception of the ref_artifact parameter, as .zarr-based AnnData objects do not require a Reference Specification for Zarr interoperability.

vc = vit.VitessceConfig(
    schema_version="1.0.17",
    description=adata_zarr_artifact.description,
)
dataset = vc.add_dataset(name="Habib 2017").add_object(
    vit.AnnDataWrapper(
        adata_artifact=adata_zarr_artifact,
        obs_feature_matrix_path="X",
        obs_embedding_paths=["obsm/X_umap"],
        obs_embedding_names=["UMAP"],
        obs_set_paths=["obs/CellType"],
        obs_set_names=["Cell Type"],
    )
)
obs_sets = vc.add_view(vit.Component.OBS_SETS, dataset=dataset)
obs_sets_sizes = vc.add_view(vit.Component.OBS_SET_SIZES, dataset=dataset)
scatterplot = vc.add_view(vit.Component.SCATTERPLOT, dataset=dataset, mapping="UMAP")
heatmap = vc.add_view(vit.Component.HEATMAP, dataset=dataset)
genes = vc.add_view(vit.Component.FEATURE_LIST, dataset=dataset)
vc.link_views([scatterplot, heatmap], ["featureValueColormapRange"], [[0.0, 0.1]])
vc.layout(((scatterplot | obs_sets) / heatmap) | (obs_sets_sizes / genes));

Save the VitessceConfig object.

adata_zarr_vc_artifact = ln.integrations.save_vitessce_config(
    vc, description="View Habib17 in Vitessce",
)

adata_zarr_vc_artifact.view_lineage()

_images/632e3859ec1e0a3d36f01921cc1a5e6052ac275504303c192ae1402c6e830fdb.svg

Note

After running save_vitessce_config, a Vitessce button will appear next to the dataset in the Artifacts page of the web interface.

Visualize a SpatialData object (Zarr format)¶

Download and unzip the example SpatialData dataset (approximately 1.47 GB unzipped). Once unzipped, the data will be available via a local Zarr DirectoryStore.

sdata_zip_filepath = "./visium.spatialdata.zarr.zip"
sdata_zarr_filepath = "./visium.spatialdata.zarr"
if not Path(sdata_zarr_filepath).exists():
    if not Path(sdata_zip_filepath).exists():
        urlretrieve(
            "https://s3.embl.de/spatialdata/spatialdata-sandbox/visium_associated_xenium_io.zip",
            sdata_zip_filepath,
        )
    with zipfile.ZipFile(sdata_zip_filepath, "r") as zip_ref:
        zip_ref.extractall(".")
        Path("data.zarr").rename(sdata_zarr_filepath)

Save the dataset¶

Save the .zarr DirectoryStore as a LaminDB Artifact.

sdata_zarr_artifact = ln.Artifact(
    sdata_zarr_filepath,
    key="vitessce_examples/visium.sdata.zarr",
    description="Visium SpatialData Example",
    type="dataset",
)
sdata_zarr_artifact.save()

! record with similar description exists! did you mean to load it?

	uid	key	description	suffix	type	size	hash	n_objects	n_observations	_hash_type	_accessor	visibility	_key_is_virtual	storage_id	transform_id	version	is_latest	run_id	created_at	created_by_id
id
2934	TqvTDISGLCT5TgWA0000	None	View Visium SpatialData Example in Vitessce	.vitessce.json	None	1847	dDQ9YoFFhK6ff9PW2IXfGQ	None	None	md5	None	1	True	1	108	None	True	382	2024-12-20 15:08:08.128741+00:00	2

Artifact(uid='7KxqCuqizBwVaAp10000', is_latest=True, key='vitessce_examples/visium.sdata.zarr', description='Visium SpatialData Example', suffix='.zarr', type='dataset', size=1473180203, hash='DZDiAPrxyh9g8dOn0HmIfQ', n_objects=9167, _hash_type='md5-d', visibility=1, _key_is_virtual=True, storage_id=1, transform_id=126, run_id=246, created_by_id=2, created_at=2024-12-20 15:41:48 UTC)

Save a VitessceConfig object¶

You can create a dashboard for one or several datasets by using Vitessce’s component API. Here, we use the SpatialDataWrapper class to specify which parts of the SpatialData object will be loaded for visualization.

vc = vit.VitessceConfig(
    schema_version="1.0.17",
    description=sdata_zarr_artifact.description,
)
# Add data to the configuration:
dataset_uid = "sdata_visium"
dataset = vc.add_dataset(name='Breast Cancer Visium', uid=dataset_uid).add_object(
    vit.SpatialDataWrapper(
        sdata_artifact=sdata_zarr_artifact,
        # The following paths are relative to the root of the SpatialData zarr store on-disk.
        image_path="images/CytAssist_FFPE_Human_Breast_Cancer_full_image",
        table_path="tables/table",
        obs_feature_matrix_path="tables/table/X",
        obs_spots_path="shapes/CytAssist_FFPE_Human_Breast_Cancer",
        region="CytAssist_FFPE_Human_Breast_Cancer",
        coordinate_system="global",
        coordination_values={
            # The following tells Vitessce to consider each observation as a "spot"
            "obsType": "spot",
        },
    )
)

# Add views (visualizations) to the configuration:
spatial = vc.add_view("spatialBeta", dataset=dataset)
feature_list = vc.add_view("featureList", dataset=dataset)
layer_controller = vc.add_view("layerControllerBeta", dataset=dataset)

# Initialize visual properties for multiple linked views:
vc.link_views_by_dict([spatial, layer_controller], {
    "imageLayer": vit.CoordinationLevel([{
        "photometricInterpretation": "RGB",
    }]),
}, scope_prefix=vit.get_initial_coordination_scope_prefix(dataset_uid, "image"))
vc.link_views([spatial, layer_controller, feature_list, obs_sets], ["obsType"], ["spot"])

# Layout the views
vc.layout(spatial | (feature_list / layer_controller));

Save the VitessceConfig object.

sdata_vc_artifact = ln.integrations.save_vitessce_config(
    vc, description="View Visium SpatialData Example in Vitessce",
)

sdata_vc_artifact.view_lineage()

_images/89a16815ea3d4b861db6ff2d971f28d85aff703a6e5aad1ad140d603aee96797.svg

Note

After running save_vitessce_config, a Vitessce button will appear next to the dataset in the Artifacts page of the web interface.

Visualize an image (OME-TIFF format)¶

Vitesse can visualize data from multiple bioimaging file formats, including OME-TIFF.

Here, we download an example OME-TIFF file (approximately 1.02 GB).

ome_tiff_filepath = "./VAN0006-LK-2-85-PAS_registered.ome.tif"
if not Path(ome_tiff_filepath).exists():
    urlretrieve(
        "https://assets.hubmapconsortium.org/f4188a148e4c759092d19369d310883b/ometiff-pyramids/processedMicroscopy/VAN0006-LK-2-85-PAS_images/VAN0006-LK-2-85-PAS_registered.ome.tif",
        ome_tiff_filepath,
    )

Save the dataset¶

Save the .ome.tif file as a LaminDB Artifact.

ome_tiff_artifact = ln.Artifact(
    ome_tiff_filepath,
    key="vitessce_examples/VAN0006-LK-2-85-PAS_registered.ome.tif",
    description="PAS OME-TIFF file, Neumann et al., 2020",
    type="dataset",
)
ome_tiff_artifact.save()

Artifact(uid='PFiIwkDJOAsn6lDq0000', is_latest=True, key='vitessce_examples/VAN0006-LK-2-85-PAS_registered.ome.tif', description='PAS OME-TIFF file, Neumann et al., 2020', suffix='.tif', type='dataset', size=1021561156, hash='pa3FYqKL6afG8PduprhRJk', _hash_type='sha1-fl', visibility=1, _key_is_virtual=True, storage_id=1, transform_id=126, run_id=246, created_by_id=2, created_at=2024-12-20 15:47:48 UTC)

When using OME-TIFF files, we can use generate-tiff-offsets to create an index for the bytes within the OME-TIFF file. We store these to a companion .offsets.json file which makes loading subsets of the image more efficient.

offsets = get_offsets(ome_tiff_filepath)

offsets_path = "./VAN0006-LK-2-85-PAS_registered.offsets.json"
with open(offsets_path, "w") as file:
    json.dump(offsets, file)
offsets_artifact = ln.Artifact(
    offsets_path,
    key="vitessce_examples/VAN0006-LK-2-85-PAS_registered.offsets.json",
    description="PAS offsets file, Neumann et al., 2020"
).save()

Save a VitessceConfig object¶

You can create a dashboard for one or several datasets by using Vitessce’s component API. Here, we use the ImageOmeTiffWrapper class to specify which pair of OME-TIFF file and offsets JSON file to load.

vc = vit.VitessceConfig(
    schema_version="1.0.17",
    description=ome_tiff_artifact.description
)
dataset = vc.add_dataset("Image").add_object(
    vit.ImageOmeTiffWrapper(
        img_artifact=ome_tiff_artifact,
        offsets_artifact=offsets_artifact,
    )
)
spatial = vc.add_view("spatialBeta", dataset=dataset)
layer_controller = vc.add_view("layerControllerBeta", dataset=dataset)
vc.layout(spatial | layer_controller);

Save the VitessceConfig object.

description_pas_ome_tiff = "View PAS OME-TIFF, Neumann et al., 2020 in Vitessce"
ome_tiff_vc_artifact = ln.integrations.save_vitessce_config(
    vc, description=description_pas_ome_tiff,
)

ome_tiff_vc_artifact.view_lineage()

_images/c02eed1699b02e9ed61969cb8c62a9459242817a48b4abae3dcb9a693a10a8cc.svg

Note

After running save_vitessce_config, a Vitessce button will appear next to the dataset in the Artifacts page of the web interface.

If your VitessceConfig object references data from multiple artifacts, the Vitessce button will appear next to a Collection that groups these artifacts (on the Collections tab of the Artifacts page). In the case of OME-TIFF, the presence of the corresponding offsets JSON file will result in the creation of a Collection.

Visualize an image (OME-Zarr format)¶

Retrieve an OME-Zarr (also known as OME-NGFF) formatted image. We could download an unzip such an image, but in downloading the SpatialData object above, we already have access to a valid OME-Zarr image locally.

ome_zarr_filepath = Path(sdata_zarr_filepath) / "images" / "CytAssist_FFPE_Human_Breast_Cancer_full_image"

Save the dataset¶

Save the .ome.zarr DirectoryStore as a LaminDB Artifact.

ome_zarr_artifact = ln.Artifact(
    ome_zarr_filepath,
    key="vitessce_examples/visium.ome.zarr",
    description="Visium OME-Zarr Example",
    type="dataset",
)
ome_zarr_artifact.save()

Artifact(uid='yP5MfDXiDNqlryqv0000', is_latest=True, key='vitessce_examples/visium.ome.zarr', description='Visium OME-Zarr Example', suffix='', type='dataset', size=1400251840, hash='UtsNOkk05EFL34hgUabEwg', n_objects=8809, _hash_type='md5-d', visibility=1, _key_is_virtual=True, storage_id=1, transform_id=126, run_id=246, created_by_id=2, created_at=2024-12-20 15:48:39 UTC)

Save a VitessceConfig object¶

You can create a dashboard for one or several datasets by using Vitessce’s component API. Here, we use the ImageOmeZarrWrapper class to specify an OME-Zarr file to load for visualization.

vc = vit.VitessceConfig(
    schema_version="1.0.17",
    description=ome_zarr_artifact.description
)
dataset_uid = "ome_zarr_image"
dataset = vc.add_dataset("Image", uid=dataset_uid).add_object(
    vit.ImageOmeZarrWrapper(
        img_artifact=ome_zarr_artifact,
    )
)
spatial = vc.add_view("spatialBeta", dataset=dataset)
layer_controller = vc.add_view("layerControllerBeta", dataset=dataset)
vc.link_views_by_dict([spatial, layer_controller], {
    "imageLayer": vit.CoordinationLevel([{
        "photometricInterpretation": "RGB",
    }]),
}, scope_prefix=vit.get_initial_coordination_scope_prefix(dataset_uid, "image"))
vc.layout(spatial | layer_controller);

Save the VitessceConfig object.

ome_zarr_vc_artifact = ln.integrations.save_vitessce_config(
    vc, description="View Visium OME-Zarr Example in Vitessce",
)

ome_zarr_vc_artifact.view_lineage()

_images/f7a1a341882a886a9ca7af30c6457deb9081480b7761e0cc71d76b3081c9b34f.svg

Note

After running save_vitessce_config, a Vitessce button will appear next to the dataset in the Artifacts page of the web interface.

# [optional] finish run and auto-save the notebook
# ln.finish()