From 39e8284310227eaafcbb8e14139a1d03763c304c Mon Sep 17 00:00:00 2001
From: NeuralFlux <40491005+NeuralFlux@users.noreply.github.com>
Date: Tue, 25 Jun 2024 13:01:50 -0400
Subject: [PATCH] added basic plugin for new source gtopdb

---
 src/plugins/gtopdb/manifest.yaml |  17 +++++
 src/plugins/gtopdb/parser.py     | 122 +++++++++++++++++++++++++++++++
 2 files changed, 139 insertions(+)
 create mode 100644 src/plugins/gtopdb/manifest.yaml
 create mode 100644 src/plugins/gtopdb/parser.py

diff --git a/src/plugins/gtopdb/manifest.yaml b/src/plugins/gtopdb/manifest.yaml
new file mode 100644
index 0000000..1382a89
--- /dev/null
+++ b/src/plugins/gtopdb/manifest.yaml
@@ -0,0 +1,17 @@
+version: '0.2'
+requires:
+  - "numpy==1.26.4"
+  - "pandas==2.2.2"
+  - "lxml==5.2.2"
+__metadata__:
+  license_url: https://www.guidetopharmacology.org/download.jsp
+  licence: CC BY-SA 4.0
+  url: https://www.guidetopharmacology.org
+dumper:
+  data_url:
+    - https://www.guidetopharmacology.org/DATA/approved_drug_detailed_interactions.csv
+    - https://www.guidetopharmacology.org/DATA/ligands.csv
+  uncompress: false
+uploader:
+  parser: parser:load_ligands
+  on_duplicates: ignore
diff --git a/src/plugins/gtopdb/parser.py b/src/plugins/gtopdb/parser.py
new file mode 100644
index 0000000..208b5d3
--- /dev/null
+++ b/src/plugins/gtopdb/parser.py
@@ -0,0 +1,122 @@
+import os
+
+import lxml.html
+import numpy as np
+import pandas as pd
+from biothings import config
+from biothings.utils.dataload import dict_convert, dict_sweep
+
+logging = config.logger
+
+VAL_MAP = {"yes": True, "no": False}
+process_key = lambda key: key.replace(" ", "_").lower()
+process_val = lambda val: VAL_MAP[val] if isinstance(val, str) and val in VAL_MAP.keys() else val
+remove_tags = lambda val: (
+    lxml.html.document_fromstring(val).text_content() if isinstance(val, str) else val
+)
+intrs_rename_dict = {
+    "Target Ensembl Gene ID": "Ensembl Gene",
+    "Target Entrez Gene ID": "Entrez Gene",
+    "Target Gene Name": "Symbol",
+    "Target Species": "Species",
+}
+
+
+def preprocess_ligands(d: dict):
+    """convert key names, remove empty vals and XML tags, and determine _id
+
+    Args:
+        d (dict): ligand properties
+
+    Returns:
+        dict: processed ligand properties
+    """
+    if isinstance(d["Synonyms"], str):
+        d["Synonyms"] = d["Synonyms"].split("|")
+    d = dict_sweep(d, vals=["", np.nan], remove_invalid_list=True)
+    d = dict_convert(d, keyfn=process_key)
+    d = dict_convert(d, valuefn=process_val)
+    d = dict_convert(d, valuefn=remove_tags)
+
+    if "inchikey" in d.keys() and not d["inchikey_dup"]:
+        d["_id"] = d["inchikey"]
+    elif "pubchem_cid" in d.keys() and not d["cid_dup"]:
+        d["_id"] = f"pubchem.compound:{d['pubchem_cid']}"
+    elif "pubchem_sid" in d.keys() and not d["sid_dup"]:
+        d["_id"] = f"pubchem.substance:{d['pubchem_sid']}"
+
+    for key in ["inchikey_dup", "cid_dup", "sid_dup"]:
+        d.pop(key)
+    return d
+
+
+def preprocess_intrs(d: dict):
+    """convert key names and remove empty vals, XML tags, and repeated columns
+
+    Args:
+        d (dict): interaction properties
+
+    Returns:
+        dict: processed interaction properties
+    """
+    d["Name"] = d["Target"]
+    if isinstance(d["Species"], str):
+        d["Species"] = d["Species"].lower()
+
+    # redundant since present in ligands
+    cols_to_drop = [
+        "Ligand ID",
+        "CAS Number",
+        "Clinical Use Comment",
+        "Bioactivity Comment",
+        "Ligand Synonyms",
+        "Target",
+        "Ligand",
+        "Type",
+        "SMILES",
+    ]
+    for col in cols_to_drop:
+        d.pop(col)
+
+    d = dict_sweep(d, vals=["", np.nan], remove_invalid_list=True)
+    d = dict_convert(d, keyfn=process_key)
+    d = dict_convert(d, valuefn=remove_tags)
+    return d
+
+
+def load_ligands(data_folder: str):
+    # pk: Ligand ID,Target ID,Target Ligand ID,Target Species
+    # inner join of primary_targets_csv[pk] and detailed_csv[pk] is primary_targets_csv[pk]
+    interactions_file = os.path.join(data_folder, "approved_drug_detailed_interactions.csv")
+    ligands_file = os.path.join(data_folder, "ligands.csv")
+    assert os.path.exists(interactions_file) and os.path.exists(ligands_file)
+
+    ligands = pd.read_csv(ligands_file, skiprows=1, dtype=object).set_index("Ligand ID")
+    interactions = (
+        pd.read_csv(interactions_file, skiprows=1, dtype=object)
+        .rename(intrs_rename_dict, axis=1)
+        .to_dict(orient="records")
+    )
+    assert type(list(ligands.keys())[0]) == str
+
+    # keep track of duplicates for determining _id
+    ligands["inchikey_dup"] = ligands["InChIKey"].duplicated()
+    ligands["cid_dup"] = ligands["PubChem CID"].duplicated()
+    ligands["sid_dup"] = ligands["PubChem SID"].duplicated()
+    ligands = ligands.to_dict(orient="index")
+
+    for row in interactions:
+        ligand_id = str(row["Ligand ID"])
+
+        # NOTE: we assume ligand IDs in interactions will be found in ligands
+        if "interaction_targets" not in ligands[ligand_id].keys():
+            ligands[ligand_id]["interaction_targets"] = []
+            ligands[ligand_id]["CAS Number"] = row["CAS Number"]
+            ligands[ligand_id]["Clinical Use Comment"] = row["Clinical Use Comment"]
+            ligands[ligand_id]["Bioactivity Comment"] = row["Bioactivity Comment"]
+
+        ligands[ligand_id]["interaction_targets"].append(preprocess_intrs(row))
+
+    for k, ligand in ligands.items():
+        ligand["_id"] = f"gtopdb:{k}"  # default _id if others are NaN or duplicated
+        yield preprocess_ligands(ligand)