CHAFF_processing_scripts/st4_smiles_curation.py

import os
import pandas as pd
import rdkit
import molvs
import tqdm
import glob
from rdkit import Chem

standardizer = molvs.Standardizer()
fragment_remover = molvs.fragment.FragmentRemover()


excluded_aids = {"584", "585", "1478", "1476", "485294", "485341"}  # with/without detergents

file_path = "./active/*.csv"

for file in glob.glob(file_path):
    file_name = os.path.splitext(os.path.basename(file))[0]
    aid = file_name.replace("pubchem_aid_", "").replace("_active", "")

    if aid in excluded_aids:
        print(f"Skipping {aid} (excluded).")
        continue

    output_path = f"./curated/{file_name}_curated.csv"
    if os.path.exists(output_path):
        print(f"Skipping {file_name}, already exists.")
        continue

    active_df = pd.read_csv(file)
    smiles_series = active_df["CanonicalSMILES"]
    active_df["curated_SMILES"] = None
    cid = active_df["CID"]

    # --------- SMILES sanitization --------- 
    valid_indices = []
    invalid_smiles = []
    warning_smiles = []

    for idx, smiles in smiles_series.items():
        mol = Chem.MolFromSmiles(smiles)
        compound_cid = cid.iloc[idx]

        if mol is None:
            invalid_smiles.append({
                'CID': compound_cid,
                'SMILES': smiles,
                'Reason': "MolFromSmiles returned None"
            })
            continue

        results = molvs.validate_smiles(smiles)

        if len(results) > 0:
            warning_smiles.append({
                'CID': compound_cid,
                'SMILES': smiles,
                'Reason': results
            })
            continue

        mol = standardizer.standardize(mol)
        mol = fragment_remover.remove(mol)
        standardized = Chem.MolToSmiles(mol)

        active_df.at[idx, "curated_SMILES"] = standardized
        valid_indices.append(idx)

    # Save valid entries
    valid_df = active_df.loc[valid_indices].reset_index(drop=True)
    valid_df = valid_df.drop(columns=["CanonicalSMILES"])
    valid_df = valid_df.rename(columns={"curated_SMILES": "SMILES"})

    # Create DataFrames for invalid and warning
    invalid_df = pd.DataFrame(invalid_smiles)
    warning_df = pd.DataFrame(warning_smiles)

    # Add AID column to dataframes
    valid_df.insert(0, "AID", aid)
    invalid_df.insert(0, "AID", aid)
    warning_df.insert(0, "AID", aid)

    # Save csv files
    valid_df.to_csv(f'./curated/{file_name}_curated.csv', index=False)
    invalid_df.to_csv(f'./curated/{file_name}_invalid_smiles.csv', index=False)
    warning_df.to_csv(f'./curated/{file_name}_molvs_validation.csv', index=False)

    print(f"Number of compounds in {file_name}:", len(active_df))
    print(f"Number of invalid smiles in {file_name}: {len(invalid_df)}")
    print(f"Number of warning smiles in {file_name}: {len(warning_df)}\n")