Integrate scRNA-seq datasets#
scRNA-seq data integration is the process of combining and analyzing data from several scRNA sequencing experiments to uncover common or distinct biological insights and patterns.
Here, we’ll demonstrate how to fetch two scRNA-seq datasets by registered metadata such as cell types to finally integrate them.
Setup#
!lamin load test-scrna
Show code cell output
💡 found cached instance metadata: /home/runner/.lamin/instance--testuser1--test-scrna.env
✅ loaded instance: testuser1/test-scrna
import lamindb as ln
import lnschema_bionty as lb
import pandas as pd
import anndata as ad
✅ loaded instance: testuser1/test-scrna (lamindb 0.51.0)
ln.track()
💡 notebook imports: anndata==0.9.2 lamindb==0.51.0 lnschema_bionty==0.30.0 pandas==1.5.3
✅ saved: Transform(id='agayZTonayqAz8', name='Integrate scRNA-seq datasets', short_name='scrna2', version='0', type=notebook, updated_at=2023-08-28 18:24:59, created_by_id='DzTjkKse')
✅ saved: Run(id='R3YaKMp4bwPSVFk6GjsX', run_at=2023-08-28 18:24:59, transform_id='agayZTonayqAz8', created_by_id='DzTjkKse')
Query files based on metadata#
assays = lb.ExperimentalFactor.lookup()
species = lb.Species.lookup()
query = ln.File.filter(
experimental_factors=assays.single_cell_rna_sequencing, # scRNA-seq
species=species.human, # human
cell_types__name__contains="monocyte", # monocyte
).distinct()
query.df()
| storage_id | key | suffix | accessor | description | version | initial_version_id | size | hash | hash_type | transform_id | run_id | updated_at | created_by_id | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| id | ||||||||||||||
| L5bU7hgOWmZM2VCa5A5I | AXBBarQq | None | .h5ad | AnnData | 10x reference pbmc68k | None | None | 589484 | eKVXV5okt5YRYjySMTKGEw | md5 | Nv48yAceNSh8z8 | CHafmKbEL0RBBfWOoz0J | 2023-08-28 18:24:53 | DzTjkKse |
| ZYlkASClrCMvJB45qd7X | AXBBarQq | None | .h5ad | AnnData | Conde22 | None | None | 28049505 | WEFcMZxJNmMiUOFrcSTaig | md5 | Nv48yAceNSh8z8 | CHafmKbEL0RBBfWOoz0J | 2023-08-28 18:24:42 | DzTjkKse |
Intersect measured genes between two datasets#
# get file objects
file1, file2 = query.list()
file1.describe()
💡 File(id='L5bU7hgOWmZM2VCa5A5I', key=None, suffix='.h5ad', accessor='AnnData', description='10x reference pbmc68k', version=None, size=589484, hash='eKVXV5okt5YRYjySMTKGEw', hash_type='md5', created_at=2023-08-28 18:24:53, updated_at=2023-08-28 18:24:53)
Provenance:
🗃️ storage: Storage(id='AXBBarQq', root='/home/runner/work/lamin-usecases/lamin-usecases/docs/test-scrna', type='local', updated_at=2023-08-28 18:24:58, created_by_id='DzTjkKse')
📔 transform: Transform(id='Nv48yAceNSh8z8', name='Validate & register scRNA-seq datasets', short_name='scrna', version='0', type='notebook', updated_at=2023-08-28 18:24:53, created_by_id='DzTjkKse')
👣 run: Run(id='CHafmKbEL0RBBfWOoz0J', run_at=2023-08-28 18:24:03, transform_id='Nv48yAceNSh8z8', created_by_id='DzTjkKse')
👤 created_by: User(id='DzTjkKse', handle='testuser1', email='testuser1@lamin.ai', name='Test User1', updated_at=2023-08-28 18:24:58)
Features:
var (X):
🔗 index (695, bionty.Gene.id): ['SkU86c6dfr0B', 'wUYYqQhHOQ1T', 'FJ4p0HleLknx', 'mLZxpATriwGh', '54Nah0TGq83q'...]
external:
🔗 assay (1, bionty.ExperimentalFactor): ['single-cell RNA sequencing']
🔗 species (1, bionty.Species): ['human']
obs (metadata):
🔗 cell_type (9, bionty.CellType): ['CD14-positive, CD16-negative classical monocyte', 'CD16-positive, CD56-dim natural killer cell, human', 'cytotoxic T cell', 'B cell, CD19-positive', 'effector memory CD4-positive, alpha-beta T cell, terminally differentiated']
file1.view_lineage()
file2.describe()
💡 File(id='ZYlkASClrCMvJB45qd7X', key=None, suffix='.h5ad', accessor='AnnData', description='Conde22', version=None, size=28049505, hash='WEFcMZxJNmMiUOFrcSTaig', hash_type='md5', created_at=2023-08-28 18:24:42, updated_at=2023-08-28 18:24:42)
Provenance:
🗃️ storage: Storage(id='AXBBarQq', root='/home/runner/work/lamin-usecases/lamin-usecases/docs/test-scrna', type='local', updated_at=2023-08-28 18:24:58, created_by_id='DzTjkKse')
📔 transform: Transform(id='Nv48yAceNSh8z8', name='Validate & register scRNA-seq datasets', short_name='scrna', version='0', type='notebook', updated_at=2023-08-28 18:24:53, created_by_id='DzTjkKse')
👣 run: Run(id='CHafmKbEL0RBBfWOoz0J', run_at=2023-08-28 18:24:03, transform_id='Nv48yAceNSh8z8', created_by_id='DzTjkKse')
👤 created_by: User(id='DzTjkKse', handle='testuser1', email='testuser1@lamin.ai', name='Test User1', updated_at=2023-08-28 18:24:58)
Features:
var (X):
🔗 index (36390, bionty.Gene.id): ['nhFjFmu1xBZI', 'SqsX0a250Sys', 'mkHhz0ai3yXT', 'NA1yysqkE6PQ', 'imiSu6lCPdLv'...]
obs (metadata):
🔗 cell_type (32, bionty.CellType): ['CD16-positive, CD56-dim natural killer cell, human', 'macrophage', 'mucosal invariant T cell', 'naive thymus-derived CD4-positive, alpha-beta T cell', 'plasmablast']
🔗 assay (4, bionty.ExperimentalFactor): ["10x 5' v2", "10x 5' v1", "10x 3' v3", 'single-cell RNA sequencing']
🔗 tissue (17, bionty.Tissue): ['thoracic lymph node', 'omentum', 'skeletal muscle tissue', 'blood', 'jejunal epithelium']
🔗 donor (12, core.Label): ['A52', 'A35', '637C', 'A36', 'D503']
file2.view_lineage()
Load files into memory:
file1_adata = file1.load()
file2_adata = file2.load()
💡 adding file L5bU7hgOWmZM2VCa5A5I as input for run R3YaKMp4bwPSVFk6GjsX, adding parent transform Nv48yAceNSh8z8
💡 adding file ZYlkASClrCMvJB45qd7X as input for run R3YaKMp4bwPSVFk6GjsX, adding parent transform Nv48yAceNSh8z8
Here we compute shared genes without loading files:
file1_genes = file1.features["var"]
file2_genes = file2.features["var"]
shared_genes = file1_genes & file2_genes
len(shared_genes)
695
shared_genes.list("symbol")[:10]
['SUPT4H1',
'FGR',
'PSMB6',
'NUDCD2',
'SPINT2',
'DNAJB1',
'EAF2',
'ANKRD12',
'LY6E',
'GABARAPL2']
We also need to convert the ensembl_gene_id to symbol for file2 so that they can be concatenated:
mapper = pd.DataFrame(shared_genes.values_list("ensembl_gene_id", "symbol")).set_index(
0
)[1]
mapper.head()
0
ENSG00000213246 SUPT4H1
ENSG00000000938 FGR
ENSG00000142507 PSMB6
ENSG00000170584 NUDCD2
ENSG00000167642 SPINT2
Name: 1, dtype: object
file2_adata.var.rename(index=mapper, inplace=True)
Intersect cell types#
file1_celltypes = file1.cell_types.all()
file2_celltypes = file2.cell_types.all()
shared_celltypes = file1_celltypes & file2_celltypes
shared_celltypes_names = shared_celltypes.list("name")
shared_celltypes_names
['CD16-positive, CD56-dim natural killer cell, human',
'conventional dendritic cell']
We can now subset the two datasets by shared cell types:
file1_adata_subset = file1_adata[
file1_adata.obs["cell_type"].isin(shared_celltypes_names)
]
file2_adata_subset = file2_adata[
file2_adata.obs["cell_type"].isin(shared_celltypes_names)
]
Concatenate subseted datasets:
adata_concat = ad.concat(
[file1_adata_subset, file2_adata_subset],
label="file",
keys=[file1.description, file2.description],
)
adata_concat
AnnData object with n_obs × n_vars = 126 × 695
obs: 'cell_type', 'file'
obsm: 'X_umap'
adata_concat.obs.value_counts()
cell_type file
CD16-positive, CD56-dim natural killer cell, human Conde22 114
conventional dendritic cell Conde22 7
CD16-positive, CD56-dim natural killer cell, human 10x reference pbmc68k 3
conventional dendritic cell 10x reference pbmc68k 2
dtype: int64
Show code cell content
# clean up test instance
!lamin delete --force test-scrna
!rm -r ./test-scrna
💡 deleting instance testuser1/test-scrna
✅ deleted instance settings file: /home/runner/.lamin/instance--testuser1--test-scrna.env
✅ instance cache deleted
✅ deleted '.lndb' sqlite file
❗ consider manually deleting your stored data: /home/runner/work/lamin-usecases/lamin-usecases/docs/test-scrna