CellMarker¶
lamindb provides access to the following public cell marker ontologies through bionty:
Here we show how to access and search cell marker ontologies to standardize new data.
import bionty as bt
import pandas as pd
PublicOntology objects¶
Let us create a public ontology accessor with public(), which chooses a default public ontology source from Source. It’s a PublicOntology object, which you can think about as a public registry:
public = bt.CellMarker.public(organism="human")
public
→ connected lamindb: testuser1/test-public-ontologies
PublicOntology
Entity: CellMarker
Organism: human
Source: cellmarker, 2.0
#terms: 15466
As for registries, you can export the ontology as a DataFrame:
df = public.df()
df.head()
| name | synonyms | gene_symbol | ncbi_gene_id | uniprotkb_id | |
|---|---|---|---|---|---|
| 0 | A1BG | A1BG | 1 | P04217 | |
| 1 | A2M | A2M | 3494 | None | |
| 2 | A2ML1 | A2ML1 | 144568 | A8K2U0 | |
| 3 | A4GALT | A4GALT | 53947 | A0A0S2Z5J1 | |
| 4 | AADAC | AADAC | 13 | P22760 |
Unlike registries, you can also export it as a Pronto object via public.ontology.
Look up terms¶
As for registries, terms can be looked up with auto-complete:
lookup = public.lookup()
The . accessor provides normalized terms (lower case, only contains alphanumeric characters and underscores):
lookup.immp1l
CellMarker(name='IMMP1L', synonyms='', gene_symbol='IMMP1L', ncbi_gene_id='196294', uniprotkb_id='Q96LU5')
To look up the exact original strings, convert the lookup object to dict and use the [] accessor:
lookup_dict = lookup.dict()
lookup_dict["IMMP1L"]
CellMarker(name='IMMP1L', synonyms='', gene_symbol='IMMP1L', ncbi_gene_id='196294', uniprotkb_id='Q96LU5')
Search terms¶
Search behaves in the same way as it does for registries:
public.search("CD4").head(5)
| name | synonyms | gene_symbol | ncbi_gene_id | uniprotkb_id | |
|---|---|---|---|---|---|
| 1900 | Cd4 | CD4 | 920 | B4DT49 | |
| 1901 | CD40 | CD40 | 958 | A0A0S2Z3C7 | |
| 1905 | CD40LG | CD40LG | 959 | P29965 | |
| 1908 | CD46 | CD46 | 4179 | P15529 | |
| 1907 | Cd44 | CD44 | 960 | P16070 |
Search another field (default is .name):
public.search("CD4", field=public.gene_symbol).head(1)
| name | synonyms | gene_symbol | ncbi_gene_id | uniprotkb_id | |
|---|---|---|---|---|---|
| 1900 | Cd4 | CD4 | 920 | B4DT49 |
Standardize cell marker identifiers¶
Let us generate a DataFrame that stores a number of cell markers identifiers, some of which corrupted:
markers = pd.DataFrame(
index=[
"KI67",
"CCR7",
"CD14",
"CD8",
"CD45RA",
"CD4",
"CD3",
"CD127a",
"PD1",
"Invalid-1",
"Invalid-2",
"CD66b",
"Siglec8",
"Time",
]
)
Now let’s check which cell markers can be found in the reference:
public.inspect(markers.index, public.name);
! 8 unique terms (57.10%) are not validated for name: 'KI67', 'CCR7', 'CD14', 'CD4', 'CD127a', 'Invalid-1', 'Invalid-2', 'Time'
detected 4 unique terms with inconsistent casing/synonyms: KI67, CCR7, CD14, CD4
→ standardize terms via .standardize()
Logging suggests to map synonyms:
synonyms_mapper = public.standardize(markers.index, return_mapper=True)
synonyms_mapper
{'KI67': 'Ki67', 'CCR7': 'Ccr7', 'CD14': 'Cd14', 'CD4': 'Cd4'}
Let’s replace the synonyms with standardized names in the DataFrame:
markers.rename(index=synonyms_mapper, inplace=True)
The Time, Invalid-1 and Invalid-2 are non-marker channels which won’t be curated by cell marker:
public.inspect(markers.index, public.name);
! 4 unique terms (28.60%) are not validated for name: 'CD127a', 'Invalid-1', 'Invalid-2', 'Time'
We don’t find CD127a, let’s check in the lookup with auto-completion:
lookup = public.lookup()
lookup.cd127
CellMarker(name='CD127', synonyms='', gene_symbol='IL7R', ncbi_gene_id='3575', uniprotkb_id='P16871', _5='cd127')
It should be cd127, we had a typo there with cd127a:
curated_df = markers.rename(index={"CD127a": lookup.cd127.name})
Optionally, search:
public.search("CD127a").head()
| name | synonyms | gene_symbol | ncbi_gene_id | uniprotkb_id | __agg__ |
|---|
Now we see that all cell marker candidates validate:
public.validate(curated_df.index, public.name);
! 3 unique terms (21.40%) are not validated: 'Invalid-1', 'Invalid-2', 'Time'
Ontology source versions¶
For any given entity, we can choose from a number of versions:
bt.Source.filter(entity="bionty.CellMarker").df()
# only lists the sources that are currently used
bt.Source.filter(entity="bionty.CellMarker", currently_used=True).df()
| uid | entity | organism | name | in_db | currently_used | description | url | md5 | source_website | space_id | dataframe_artifact_id | version | run_id | created_at | created_by_id | _aux | branch_id | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| id | ||||||||||||||||||
| 12 | 3kDh8qAX | bionty.CellMarker | human | cellmarker | False | True | CellMarker | s3://bionty-assets/human_cellmarker_2.0_CellMa... | None | http://bio-bigdata.hrbmu.edu.cn/CellMarker | 1 | None | 2.0 | None | 2025-07-14 06:41:44.843000+00:00 | 1 | None | 1 |
| 13 | 7bV5uJo3 | bionty.CellMarker | mouse | cellmarker | False | True | CellMarker | s3://bionty-assets/mouse_cellmarker_2.0_CellMa... | None | http://bio-bigdata.hrbmu.edu.cn/CellMarker | 1 | None | 2.0 | None | 2025-07-14 06:41:44.843000+00:00 | 1 | None | 1 |
When instantiating a Bionty object, we can choose a source or version:
source = bt.Source.get(name="cellmarker", version="2.0", organism="human")
public = bt.CellMarker.public(source=source)
public
PublicOntology
Entity: CellMarker
Organism: human
Source: cellmarker, 2.0
#terms: 15466
The currently used ontologies can be displayed using:
bt.Source.filter(currently_used=True).df()