Reconciling cell types
In this showcase we'll try to reconcile some cell-type data from the Tabula muris dataset.
First off, let's load the necessary packages and read in the data using Pandas. For speed and simplicity, I'll only be reconciling the unique tissue/cell_ontology_class pairs.
Pre-processing
from reconciler import reconcile
import pandas as pd
data_url = "https://s3-eu-west-1.amazonaws.com/pfigshare-u-files/10039264/annotations_droplets.csv"
cell_table = pd.read_csv(data_url)
cell_table.head()
| cell | tissue | cell_ontology_class | cell_ontology_term_iri | cell_ontology_id |
|---|---|---|---|---|
| 10X_P4_3_AAAGTAGAGATGCCAG | Bladder | mesenchymal cell | http://purl.obolibrary.org/obo/CL_0008019 | CL:0008019 |
| 10X_P4_3_AACCGCGTCCAACCAA | Bladder | mesenchymal cell | http://purl.obolibrary.org/obo/CL_0008019 | CL:0008019 |
| 10X_P4_3_AACTCCCGTCGGGTCT | Bladder | mesenchymal cell | http://purl.obolibrary.org/obo/CL_0008019 | CL:0008019 |
| 10X_P4_3_AACTCTTAGTTGCAGG | Bladder | bladder cell | http://purl.obolibrary.org/obo/CL_1001319 | CL:1001319 |
| 10X_P4_3_AACTCTTTCATAACCG | Bladder | mesenchymal cell | http://purl.obolibrary.org/obo/CL_0008019 | CL:0008019 |
Filtering only the unique pairs:
unique_cells = cell_table.drop_duplicates(subset=['tissue', 'cell_ontology_class'])
Reconciliation
Reconciling, against cell type (Q189118), returning the first 2 matches for each item:
This step will take a while to complete, varying according to your upload speed, here it took around a minute.
reconciled = reconcile(unique_cells['cell_ontology_class'], type_id="Q189118", top_res=2)
reconciled.head(10)
The output I got:
| id | match | name | score | type | type_id | input_value |
|---|---|---|---|---|---|---|
| Q1922379 | True | mesenchymal stem cells | 100 | cell type | Q189118 | mesenchymal cell |
| Q66563456 | False | epithelial cell of gall bladder | 28 | [] | nan | bladder cell |
| Q66568549 | False | urothelial cell of trigone of urinary bladder | 21 | [] | nan | bladder cell |
| Q11394395 | False | endothelial cells | 50 | [] | nan | endothelial cell |
| Q68620792 | False | human sinusoidal endothelial cell | 32.5 | [] | nan | endothelial cell |
| Q66590632 | False | basal cell of urothelium | 50 | [] | nan | basal cell of urothelium |
| Q66590636 | False | basal cell layer of urothelium | 44.5 | [] | nan | basal cell of urothelium |
| Q223143 | False | granulocyte | 67 | cell type | Q189118 | leukocyte |
| Q1775422 | False | agranulocyte | 60 | cell type | Q189118 | leukocyte |
| Q463418 | True | fibroblast | 100 | cell type | Q189118 | fibroblast |
Now if you look at the object, you will see the matches retrieved for each item.
Interestingly, as of 30 Aug. 2020, there's not a lot of cell type data present in Wikidata, a lot of the matches didn't even return a "type" value! That means they don't even have an 'instance of' property. This could be very interesting to look into, and add this information.