Manage biological knowledge#

Important

LaminDB speaks biology through Bionty schema module.

A basic wetlab schema maps R&D operations.

Let’s first review a last concept: knowledge-derived entities.

If tracked data isn’t curated against knowledge-derived standards, data integration often becomes a pain.

LaminDB with the Bionty extension can manage knowledge by providing lookups, curation and knowledge-coupled SQL tables.

[lnschema-bionty] contains the schema tables for all entities that are avilable in Bionty.

Let us map data on knowledge to illustrate the process.

import lamindb as ln
import lnschema_bionty as bt

ln.track()

ℹ️ Instance: testuser1/lnbionty-test

ℹ️ User: testuser1

ℹ️ Added notebook: Transform(id='f7F0c2n2Ft1s', v='0', name='knowledge', type=notebook, title='Manage biological knowledge', created_by='DzTjkKse', created_at=datetime.datetime(2023, 3, 30, 23, 17, 34))

ℹ️ Added run: Run(id='XE0faLyeY9bXU4hLhSlV', transform_id='f7F0c2n2Ft1s', transform_v='0', created_by='DzTjkKse', created_at=datetime.datetime(2023, 3, 30, 23, 17, 34))

Tip

You can view the reference of current biological entities in LaminDB:

ln.select(bt.dev.BiontyVersions).join(bt.dev.CurrentBiontyVersions).df()

	entity	database	database_v	database_url	created_by	created_at	updated_at
id
0	Species	ensembl	release-108	https://ftp.ensembl.org/pub/release-108/mysql/	DzTjkKse	2023-03-30 23:17:09	None
1	Gene	ensembl	release-108	https://ftp.ensembl.org/pub/release-108/mysql/	DzTjkKse	2023-03-30 23:17:09	None
2	Protein	uniprot	2022-04	https://ftp.uniprot.org/pub/databases/uniprot/...	DzTjkKse	2023-03-30 23:17:09	None
3	CellMarker	cellmarker	2.0	http://bio-bigdata.hrbmu.edu.cn/CellMarker/Cel...	DzTjkKse	2023-03-30 23:17:09	None
4	CellLine	clo	2022-03-21	https://data.bioontology.org/ontologies/CLO/su...	DzTjkKse	2023-03-30 23:17:09	None
5	CellType	cl	2023-02-15	http://purl.obolibrary.org/obo/cl/releases/202...	DzTjkKse	2023-03-30 23:17:09	None
6	Tissue	uberon	2023-02-14	http://purl.obolibrary.org/obo/uberon/releases...	DzTjkKse	2023-03-30 23:17:09	None
7	Disease	mondo	2023-02-06	http://purl.obolibrary.org/obo/mondo/releases/...	DzTjkKse	2023-03-30 23:17:09	None
8	Readout	efo	3.48.0	http://www.ebi.ac.uk/efo/releases/v3.48.0/efo.owl	DzTjkKse	2023-03-30 23:17:09	None
9	Phenotype	hp	2023-01-27	https://github.com/obophenotype/human-phenotyp...	DzTjkKse	2023-03-30 23:17:09	None
10	Pathway	pw	7.74	https://data.bioontology.org/ontologies/PW/dow...	DzTjkKse	2023-03-30 23:17:09	None

Lookup ontology ids#

For instance, you can retrieve the cell type ontology id of “gamma delta T cell” by accessing a cell type ontology through the bionty.Entity CellType:

ct_lookup = bt.CellType.lookup

ct_lookup.gamma_delta_T_cell

cell_type(ontology_id='CL:0000798', name='gamma-delta T cell')

Create knowledge-derived records#

You can also directly create a record for the CellType table:

bt.CellType(name=ct_lookup.gamma_delta_T_cell.name)

CellType(id='CL:0000798', ontology_id='CL:0000798', name='gamma-delta T cell')

Curate metadata by linking it against knowledge#

Let us link all the biological samples in a cross-tissue scRNA-seq dataset:

adata = ln.dev.datasets.anndata_human_immune_cells()
meta = adata.obs.drop_duplicates(subset=adata.obs.columns)

meta.head()

	donor_id	tissue	cell_type	assay	tissue_ontology_term_id	cell_type_ontology_term_id	assay_ontology_term_id
CZINY-0109_CTGGTCTAGTCTGTAC	D496	blood	classical monocyte	10x 3' v3	UBERON:0000178	CL:0000860	EFO:0009922
CZI-IA10244332+CZI-IA10244434_CCTTCGACATACTCTT	621B	thoracic lymph node	T follicular helper cell	10x 5' v2	UBERON:0007644	CL:0002038	EFO:0009900
Pan_T7935491_CTGGTCTGTACATGTC	A29	spleen	memory B cell	10x 5' v1	UBERON:0002106	CL:0000787	EFO:0011025
Pan_T7980367_GGGCATCCAGGTGGAT	A36	lung	alveolar macrophage	10x 5' v1	UBERON:0002048	CL:0000583	EFO:0011025
Pan_T7935494_ATCATGGTCTACCTGC	A29	mesenteric lymph node	naive thymus-derived CD4-positive, alpha-beta ...	10x 5' v1	UBERON:0002509	CL:0000895	EFO:0011025

meta.shape

(514, 7)

Let’s first add all the cell types: .curate allows you to check the passed ids are present in the knowledge table.

It returns a new DataFrame indexed with the curated ids and a boolean __curated__ column.

celltype_curate = bt.CellType.curate(meta, column="cell_type_ontology_term_id")

✅ 514 terms (100.0%) are mapped.

🔶 0 terms (0.0%) are not mapped.

Here we saw all terms can be linked. 🎉

Update the content of the knowledge-managed tables in your SQL database#

Assume there are cell types in this new dataset that are tracked in the ontology (bionty.CellType), but are not yet tracked in the DB table (lns.bionty.CellType).

Let us fix that!

We can go ahead and create records of the CellType table:

celltype_records = [bt.CellType(ontology_id=i) for i in celltype_curate.index.unique()]

celltype_records[:3]

[CellType(id='0r8seCQT', ontology_id='CL:0000860', name='classical monocyte'),
 CellType(id='mflMyS4P', ontology_id='CL:0002038', name='T follicular helper cell'),
 CellType(id='6D2sGpoW', ontology_id='CL:0000787', name='memory B cell')]

We can do the same for tissues:

tissue_curate = bt.Tissue.curate(meta, column="tissue_ontology_term_id")

✅ 514 terms (100.0%) are mapped.

🔶 0 terms (0.0%) are not mapped.

tissue_records = [bt.Tissue(ontology_id=i) for i in tissue_curate.index.unique()]

tissue_records[:3]

[Tissue(id='oOynQVrG', ontology_id='UBERON:0000178', name='blood'),
 Tissue(id='3PjYAKnU', ontology_id='UBERON:0007644', name='thoracic lymph node'),
 Tissue(id='27IniEkO', ontology_id='UBERON:0002106', name='spleen')]

Finally, let’s add them to the database:

ln.add(celltype_records + tissue_records)  # add all records in one transaction

[CellType(id='0r8seCQT', ontology_id='CL:0000860', name='classical monocyte'),
 CellType(id='mflMyS4P', ontology_id='CL:0002038', name='T follicular helper cell'),
 CellType(id='6D2sGpoW', ontology_id='CL:0000787', name='memory B cell'),
 CellType(id='mBHciTPF', ontology_id='CL:0000583', name='alveolar macrophage'),
 CellType(id='HSkf3Q1R', ontology_id='CL:0000895', name='naive thymus-derived CD4-positive, alpha-beta T cell'),
 CellType(id='4ChG7mFe', ontology_id='CL:0001062', name='effector memory CD8-positive, alpha-beta T cell, terminally differentiated'),
 CellType(id='7kDaS5Ou', ontology_id='CL:0000789', name='alpha-beta T cell'),
 CellType(id='rtDPx6Hx', ontology_id='CL:0000492', name='CD4-positive helper T cell'),
 CellType(id='9XsNkGyV', ontology_id='CL:0000900', name='naive thymus-derived CD8-positive, alpha-beta T cell'),
 CellType(id='hY2x3J1e', ontology_id='CL:0000235', name='macrophage'),
 CellType(id='0PZas9yO', ontology_id='CL:0000940', name='mucosal invariant T cell'),
 CellType(id='oXERjFK4', ontology_id='CL:0001071', name='group 3 innate lymphoid cell'),
 CellType(id='BOpRP2Fh', ontology_id='CL:0000788', name='naive B cell'),
 CellType(id='Bgz4LhGa', ontology_id='CL:0000548', name='animal cell'),
 CellType(id='TWgOU9nt', ontology_id='CL:0000938', name='CD16-negative, CD56-bright natural killer cell, human'),
 CellType(id='NJd24u0f', ontology_id='CL:0000786', name='plasma cell'),
 CellType(id='XbHAnqxj', ontology_id='CL:0000909', name='CD8-positive, alpha-beta memory T cell'),
 CellType(id='gHrHTy92', ontology_id='CL:0000939', name='CD16-positive, CD56-dim natural killer cell, human'),
 CellType(id='gVr0YFoD', ontology_id='CL:0000798', name='gamma-delta T cell'),
 CellType(id='XdmAIFHE', ontology_id='CL:0000990', name='conventional dendritic cell'),
 CellType(id='gAfMcxGY', ontology_id='CL:0001203', name='CD8-positive, alpha-beta memory T cell, CD45RO-positive'),
 CellType(id='3nr16uCT', ontology_id='CL:0000905', name='effector memory CD4-positive, alpha-beta T cell'),
 CellType(id='8NjaB9BF', ontology_id='CL:0000875', name='non-classical monocyte'),
 CellType(id='ZKZVJGmU', ontology_id='CL:0000097', name='mast cell'),
 CellType(id='s5P8y64X', ontology_id='CL:0000815', name='regulatory T cell'),
 CellType(id='8J77wB5e', ontology_id='CL:0011026', name='progenitor cell'),
 CellType(id='ZsMcdB6W', ontology_id='CL:0001056', name='dendritic cell, human'),
 CellType(id='tgTN23uh', ontology_id='CL:0000980', name='plasmablast'),
 CellType(id='j1cCEIlt', ontology_id='CL:0000784', name='plasmacytoid dendritic cell'),
 CellType(id='5H3ZzPO0', ontology_id='CL:0000542', name='lymphocyte'),
 CellType(id='4sdPZTzG', ontology_id='CL:0000844', name='germinal center B cell'),
 CellType(id='jZVwZhZ5', ontology_id='CL:0000556', name='megakaryocyte'),
 Tissue(id='oOynQVrG', ontology_id='UBERON:0000178', name='blood'),
 Tissue(id='3PjYAKnU', ontology_id='UBERON:0007644', name='thoracic lymph node'),
 Tissue(id='27IniEkO', ontology_id='UBERON:0002106', name='spleen'),
 Tissue(id='RqsZxxU5', ontology_id='UBERON:0002048', name='lung'),
 Tissue(id='oacWHW6m', ontology_id='UBERON:0002509', name='mesenteric lymph node'),
 Tissue(id='5hkBkUOD', ontology_id='UBERON:0000030', name='lamina propria'),
 Tissue(id='fDaUnIP1', ontology_id='UBERON:0002107', name='liver'),
 Tissue(id='TLoVnCfT', ontology_id='UBERON:0000400', name='jejunal epithelium'),
 Tissue(id='n2XwDJRH', ontology_id='UBERON:0003688', name='omentum'),
 Tissue(id='npJZcG4r', ontology_id='UBERON:0002371', name='bone marrow'),
 Tissue(id='ZgaNJ7qR', ontology_id='UBERON:0002116', name='ileum'),
 Tissue(id='iM008Cay', ontology_id='UBERON:0001153', name='caecum'),
 Tissue(id='nPwtH6Tb', ontology_id='UBERON:0002370', name='thymus'),
 Tissue(id='vgPMrvDI', ontology_id='UBERON:0001134', name='skeletal muscle tissue'),
 Tissue(id='MXtPqLkD', ontology_id='UBERON:0002114', name='duodenum'),
 Tissue(id='Yje0RkZC', ontology_id='UBERON:0001159', name='sigmoid colon'),
 Tissue(id='UQvkthge', ontology_id='UBERON:0001157', name='transverse colon')]

Check they are in the database:

ln.select(bt.Tissue, name="blood").one()

Tissue(id='oOynQVrG', ontology_id='UBERON:0000178', name='blood')

Now, some of the foundational knowledge is in place! 😌