EFO-OTAR mappings prototype

efo_otar_mapping.py

import rdflib
import pandas as pd

## Path to efo local or remote
owlpath = "https://github.com/EBISPOT/efo/releases/download/v3.18.0/efo_otar_slim.owl"
## File downloaded from
## https://docs.google.com/spreadsheets/d/1CV_shXJy1ACM09HZBB_-3Nl6l_dfkrA26elMAF0ttHs/edit
mappingFile = "/Users/ochoa/Downloads/OTAR project EFO mappings for disease profile pages - Sheet1-2.csv"
intraneturl = "http://home.opentargets.org/"
outputFile = "/Users/ochoa/Desktop/output.json"

## loading the graph
g = rdflib.Graph()
g.load(owlpath)

## sparql query to get all ancestors for all EFO terms
sparql_query = '''
prefix oio: <http://www.geneontology.org/formats/oboInOwl#>
prefix owl: <http://www.w3.org/2002/07/owl#>
prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#>

SELECT ?cls ?label ?ancestor ?ancestorlabel WHERE {
    ?cls a owl:Class .
    ?cls rdfs:label ?label_all .
    ?cls rdfs:subClassOf* ?ancestor .
    ?ancestor rdfs:label ?ancestorlabel_all
    BIND (STR(?ancestorlabel_all)  AS ?ancestorlabel)
    BIND (STR(?label_all)  AS ?label)
}
'''

qres = g.query(sparql_query)

## Convert into pandas data.frame
## Note: there might be better ways to searialise this
clss = []
labels = []
ancestors = []
ancestorlabels = []
for (cls, label, ancestor, ancestorlabel) in g.query(sparql_query):
    clss.append(str(cls))
    labels.append(str(label))
    ancestors.append(str(ancestor))
    ancestorlabels.append(str(ancestorlabel))

efo_df = pd.DataFrame({
    "cls": clss,
    "label": labels,
    "ancestor": ancestors,
    "ancestorlabel": ancestorlabels
})
## create ids from uri
efo_df["cls_id"] = efo_df["cls"].str.replace("(.+)/", "")
efo_df["ancestor_id"] = efo_df["ancestor"].str.replace("(.+)/", "")

## Read file with OTAR-EFO mappings
mapdf = pd.read_csv(mappingFile)

## merge 2 datasets
df = pd.merge(mapdf, efo_df, \
              how = 'inner', \
              left_on = 'EFO Disease ID',
              right_on = 'cls_id')

## Json formatting and output
df.columns = df.columns.str.strip().str.lower().str.replace(' ', '_').str.replace('(', '').str.replace(')', '')
df['reference'] = intraneturl + df['otar_code']
df.rename(columns = {'ancestor_id':'disease_id'}) \
  .groupby('disease_id') \
  .apply(lambda x: x[['otar_code','project_name','status','reference']].to_dict('r')) \
  .reset_index() \
  .rename(columns={0:'projects'}) \
  .to_json(outputFile, orient='records')

sample_output.json

  [{
    "disease_id": "Orphanet_68367",
    "projects": [
      {
        "otar_code": "OTAR2073",
        "project_name": "NeuroFlux",
        "status": "Active",
        "reference": "http://home.opentargets.org/OTAR2073"
      }
    ]
  },
  {
    "disease_id": "Orphanet_71859",
    "projects": [
      {
        "otar_code": "OTAR2072",
        "project_name": "DNA Damage",
        "status": "Active",
        "reference": "http://home.opentargets.org/OTAR2072"
      }
    ]
  }]

Pending Issue. Focusing on the ontology expansion, I have forgotten to add the direct EFO-OTAR project link to the final json file. It contains all EFO ancestors but not the EFO itself.

@dochoa here you have the code to get it properly using spark opentargets/platform-etl-backend#45
and the output generated is like this

{
  "efo_id": "MONDO_0021193",
  "projects": [
    {
      "otar_code": "OTAR041",
      "status": "Active",
      "project_name": "CELLector",
      "reference": "http://home.opentargets.org/OTAR041"
    },
    {
      "otar_code": "OTAR015",
      "status": "Active",
      "project_name": "CRISPR-Cas9 Target ID",
      "reference": "http://home.opentargets.org/OTAR015"
    },
    {
      "otar_code": "OTAR016",
      "status": "Closed",
      "project_name": "Cancer Functional Genomics",
      "reference": "http://home.opentargets.org/OTAR016"
    },
    {
      "otar_code": "OTAR2055",
      "status": "Active",
      "project_name": "Prediction of Oncology Targets and in Silico Drug Prescriptions",
      "reference": "http://home.opentargets.org/OTAR2055"
    }
  ]
}