【ChEMBL】化合物の標的タンパクを調べる

$ tar -xzvf chembl_34_postgresql.tar.gz

version: "3.6"
services:
  db:
    image: postgres:14.2
    ports:
      - 5432:5432
    tty: true
    environment:
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: password
    volumes:
      - ./db:/var/lib/postgresql/data

chembl34/chembl_34_postgresql $ docker-compose up

chembl34/chembl_34_postgresql $ mv chembl_34_postgresql.dmp db/chembl_34_postgresql.dmp

chembl34/chembl_34_postgresql $ docker container ls

chembl34/chembl_34_postgresql $ docker exec -it [docker container id] bash

(Docker内) $ pg_restore -d postgres /var/lib/postgresql/data/chembl_34_postgresql.dmp -U postgres -O

(Docker内) $ psql -d postgres -U postgres
posgre: /dt;

CREATE TABLE TEMP_COMPOUND_TARGET AS
SELECT DISTINCT
  m.chembl_id AS compound_chembl_id,
  s.canonical_smiles AS canonical_smiles,
  r.compound_name AS compound_name,
  r.cidx AS CID,
  s.standard_inchi AS inchi,
  s.standard_inchi_key AS inchi_key,
  t.chembl_id AS target_chembl_id,
  t.pref_name AS target_name,
  cs.accession AS uniprot_id
FROM compound_structures s
  RIGHT JOIN molecule_dictionary m ON s.molregno = m.molregno
  JOIN compound_records r ON m.molregno = r.molregno
  JOIN docs d ON r.doc_id = d.doc_id
  JOIN activities act ON r.record_id = act.record_id
  JOIN assays a ON act.assay_id = a.assay_id
  JOIN target_dictionary t ON a.tid = t.tid
  JOIN target_components tc ON t.tid = tc.tid
  JOIN component_sequences cs ON tc.component_id = cs.component_id
WHERE t.organism = 'Homo sapiens';

COPY TEMP_COMPOUND_TARGET TO '/var/lib/postgresql/data/TEMP_COMPOUND_TARGET.csv' DELIMITER ',' CSV HEADER;

import pandas as pd

df_compounds = pd.read_csv('chembl_34/chembl_34_postgresql/db/TEMP_COMPOUND_TARGET.csv')
df.head()

# statistics
compounds = df_compounds.inchi_key.drop_duplicates().tolist()
compounds_proteins = df_compounds.uniprot_id.drop_duplicates().tolist()
print(f'n_compounds: {len(compounds)}')
print(f'n_proteins: {len(compounds_proteins)}')

import pickle

# 1. 標的タンパクのUniProt IDリストを作成して保存
uniprotIDs = df_compounds['uniprot_id'].drop_duplicates().tolist()
f = open('id_conversion/target_protein_uniprotIDs.csv', 'w')
f.writelines(['uniprot_id\n'] + [str(i) + '\n' for i in uniprotIDs])
f.close()

# 2. RのbiomaRtで作成したUniProt IDとシンボル表記の対応表を作成して保存(uniprotID_to_symbols.csv)

# 3. 対応表を読み込んで辞書に変換し
df_unp_sym = pd.read_csv('id_conversion/uniprotID_to_symbols.csv', index_col=1)
unp_sym = df_unp_sym['uniprot_gn_symbol'].to_dict()

# 4. pickleで保存
with open('temp/uniprotID_to_symbol.pkl', 'wb') as tf:
    pickle.dump(unp_sym, tf)
    
# 5. df_compoundsにシンボルの列を加える
df_compounds['symbol'] = df_compounds['uniprot_id'].apply(
    lambda x: unp_sym[x] if x in unp_sym.keys() else np.nan
)

# 6. 化合物のInChIKeyから標的タンパクのシンボル表記を得る辞書を作成
def makeList_groupby(df, col1, col2, unique=True):
    lcol1 = df[col1].tolist()
    scol1 = list(set(lcol1))
    lcol2 = df[col2].tolist()
    
    c1_c2 = {c1:[] for c1 in scol1}
    for c1, c2 in zip(lcol1, lcol2):
        c1_c2[c1].append(c2)
        
    if unique:
        c1_c2 = {c1: list(set(c1_c2[c1])) for c1 in c1_c2.keys()}
        
    return c1_c2
  
ink_syms = makeList_groupby(df_compounds, 'inchi_key', 'symbol', unique=True)

# 7. pickleで保存
TEMP_DIR = 'temp'
with open(f'{TEMP_DIR}/compoundInChIKey_to_symbols.pkl', 'wb') as tf:
    pickle.dump(ink_syms, tf)

df_compunds.head()

with open(f'{TEMP_DIR}/pathwayTerm_to_symbols.pkl', 'rb') as tf:
    ptm_syms = pickle.load(tf)