download_enamine.py

# import requests

# class DownloadEnamine:
#     """
#     This class is set up to download Enamine REAL database on a remote machine.
#     Instatiation requires plain ``username`` and ``password``.

#     .. code-block::python
#         de = DownloadEnamine('[email protected]', 'Foo123')
#         de.download_all('REAL')

#     Note, this is copied off the route of the web page and not the Enamine Store API.
#     Plus the official documentation (emailed Word document) is for the old Store and
#     no longer applies anyway (plain text username and password in GET header "Authorization").

#     The URLs pointing to the download pages were copied off manually.
#     """
#     REAL=[
#     '2024.07_Enamine_REAL_HAC_25_1B_CXSMILES.cxsmiles.bz2',
#     ]
#     LOGIN_URL = 'https://enamine.net/compound-collections/real-compounds/real-database'

#     def __init__(self, username, password):
#         self.sesh = requests.Session()
#         login_payload = {
#             'username': username,
#             'password': password,
#             'Submit': 'Login',
#             'remember': 'yes',
#             'option': 'com_users',
#             'task': 'user.login'
#         }
#         self.sesh.headers.update({'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'})
#         response = self.sesh.post(self.LOGIN_URL, data=login_payload)
#         response.raise_for_status()

#         print("Login appears successful.")

#     def download_all(self, catalogue='REAL'):
#         """
#         The URLs of the databases files are in the class attribute of that same catalogue name (i.e. ``REAL``).
#         """
#         for filename in getattr(self, catalogue):
#             self.download('REAL', filename)

#     def check(self, catalogue='REAL'):
#         for filename in getattr(self, catalogue):
#             with self.sesh.get(f'https://ftp.enamine.net/download/{catalogue}/{filename}', stream=True) as r:
#                 r.raise_for_status()  # requests.exceptions.HTTPError
#                 for chunk in r.iter_content(chunk_size=8192):
#                     break

#     def download(self, catalogue, filename):
#         """
#         Downloads the ``filename`` of the given ``catalogue``
#         """
#         with self.sesh.get(f'https://ftp.enamine.net/download/{catalogue}/{filename}', stream=True) as r:
#             r.raise_for_status()
#             with open(filename, 'wb') as f:
#                 for chunk in r.iter_content(chunk_size=8192):
#                     f.write(chunk)


# real_download = DownloadEnamine('[email protected]', 'Z!6CJd2BjQs!y4x')
# real_download.download_all('REAL')


import os
import sys
import random # <-- Import random module

from rdkit import Chem


# --- Configuration ---
input_cxsmiles_file = "2024.07_Enamine_REAL_HAC_25_1B_CXSMILES.cxsmiles"  # <-- CHANGE THIS to your input filename
output_smiles_file = "smiles_sampled_20k.txt"     # <-- CHANGE THIS to your desired output filename

target_sample_size = 20000  # <-- How many molecules we want approximately
# IMPORTANT: Provide a reasonable estimate of the total lines in the input file.
# The filename '1B' suggests 1 billion. Accuracy affects how close the final count is to the target.
estimated_total_lines = 1_000_000_000 # <-- ADJUST if your estimate differs

# --- End Configuration ---

# --- Script Start ---

# Check if input file exists
if not os.path.isfile(input_cxsmiles_file):
    print(f"ERROR: Input file not found: {input_cxsmiles_file}")
    sys.exit(1)

# --- Calculate Sampling Probability ---
if estimated_total_lines <= 0:
    print("ERROR: estimated_total_lines must be positive.")
    sys.exit(1)

# Ensure probability is between 0 and 1
sampling_probability = min(1.0, target_sample_size / estimated_total_lines)

print(f"Reading CXSMILES from: {input_cxsmiles_file}")
print(f"Attempting to sample approximately {target_sample_size} lines.")
print(f"Based on estimated total lines: {estimated_total_lines:,}")
print(f"Calculated sampling probability per line: {sampling_probability:.8f}")
print(f"Writing standard SMILES to: {output_smiles_file}")
print("-" * 30)

count_processed_lines = 0 # Lines read from input
count_selected = 0      # Lines selected by random sampling
count_success = 0       # Selected lines successfully converted
count_error = 0         # Selected lines that failed conversion

# Open input and output files safely
try:
    with open(input_cxsmiles_file, 'r') as infile, open(output_smiles_file, 'w') as outfile:
        # Process each line in the input file
        for i, line in enumerate(infile):
            count_processed_lines += 1
            cxsmiles_line = line.strip() # Remove leading/trailing whitespace

            if not cxsmiles_line: # Skip empty lines
                continue

            # --- Random Sampling Check ---
            if random.random() < sampling_probability:
                count_selected += 1
                # --- Process Selected Line ---
                try:
                    # RDKit's MolFromSmiles often ignores CXSMILES extensions
                    # It reads the core structure.
                    mol = Chem.MolFromSmiles(cxsmiles_line)

                    if mol is not None:
                        # Convert the RDKit molecule back to a standard, canonical SMILES
                        standard_smiles = Chem.MolToSmiles(mol)
                        outfile.write(standard_smiles + '\n')
                        count_success += 1
                    else:
                        # RDKit couldn't parse this selected line
                        print(f"Warning: Could not parse selected line #{count_selected} (original line {i+1}). Input: '{cxsmiles_line}'")
                        count_error += 1

                except Exception as e:
                    # Catch any other unexpected errors during RDKit processing
                    print(f"Error processing selected line #{count_selected} (original line {i+1}): '{cxsmiles_line}'. Details: {e}")
                    count_error += 1

            # Optional: Print progress periodically for large files
            if (i + 1) % 1000000 == 0: # Print every 1 million lines processed
                 print(f"Processed {i+1:,} lines. Selected: {count_selected}. Successful conversions: {count_success}.")


except IOError as e:
    print(f"ERROR: Could not open or write file. Details: {e}")
    sys.exit(1)

print("-" * 30)
print(f"Processing finished.")
print(f"Total lines read from input: {count_processed_lines:,}")
print(f"Lines randomly selected for processing: {count_selected:,} (Target was approx. {target_sample_size:,})")
print(f"Successfully converted: {count_success:,} lines.")
print(f"Failed/Skipped (among selected): {count_error:,} lines.")
print(f"Output written to:    {output_smiles_file}")

# Add a note about potential deviation from target
if abs(count_selected - target_sample_size) > target_sample_size * 0.1: # If deviation > 10%
    print("\nNote: The actual number of selected lines differs significantly from the target.")
    print("This might be due to the 'estimated_total_lines' differing from the actual file size.")