Visualization of single-cell RNA seq data

[ ]:

import os
from os.path import join, isfile, isdir
from urllib.request import urlretrieve
from anndata import read_h5ad
import scanpy as sc
import json

from vitessce import (
    VitessceConfig,
    Component as cm,
    CoordinationType as ct,
    AnnDataWrapper,
)
from vitessce.data_utils import (
    generate_h5ad_ref_spec
)

0. Download data

[ ]:

h5_url = "https://datasets.cellxgene.cziscience.com/84df8fa1-ab53-43c9-a439-95dcb9148265.h5ad"

[ ]:

adata_filepath = join("data", "84df8fa1-ab53-43c9-a439-95dcb9148265.h5ad")
if not isfile(adata_filepath):
    os.makedirs("data", exist_ok=True)
    urlretrieve(h5_url, adata_filepath)

1. Create a Reference Spec JSON file for the H5AD file

In order for Vitessce to load H5AD files, we also need to provide a corresponding Reference Spec JSON file which contains mappings between AnnData object keys and the byte offsets at which those AnnData object values begin within the H5AD file binary contents.

[ ]:

json_filepath = join("data", "84df8fa1-ab53-43c9-a439-95dcb9148265.h5ad.reference.json")
if not isfile(json_filepath):
    ref_dict = generate_h5ad_ref_spec(h5_url)
    with open(json_filepath, "w") as f:
        json.dump(ref_dict, f)

2. Create the Vitessce widget configuration

[ ]:

vc = VitessceConfig(schema_version="1.0.17", name='Nakshatri et al', description='snRNA-seq analyses of breast tissues of healthy women of diverse genetic ancestry')

dataset = vc.add_dataset(name='84df8fa1').add_object(AnnDataWrapper(
        adata_path=adata_filepath,
        ref_path=json_filepath, # We specify paths to both the H5AD and JSON files
        obs_embedding_paths=["obsm/X_wnn.umap"],
        obs_embedding_names=["UMAP"],
        obs_set_paths=["obs/cell_type"],
        obs_set_names=["Cell Type"],
        obs_feature_matrix_path="X",
    )
)

scatterplot = vc.add_view(cm.SCATTERPLOT, dataset=dataset, mapping="UMAP")
cell_sets = vc.add_view(cm.OBS_SETS, dataset=dataset)
cell_set_sizes = vc.add_view(cm.OBS_SET_SIZES, dataset=dataset)
genes = vc.add_view(cm.FEATURE_LIST, dataset=dataset)

vc.layout((scatterplot | cell_sets) / (cell_set_sizes | genes));