scripts/st2_dataset_curation.py

# Curation of dataset
import pandas as pd
import os
from tqdm import tqdm

# Path to the input CSV and the folder containing XYZ files
csv_path = 'formed.csv'
xyz_folder = 'XYZ_FORMED'

# Load the CSV file
df = pd.read_csv(csv_path)

# Select necessary columns in 'formed.csv'
df = df[['name', 'gap']]

# Function to clean and load XYZ file data
def load_xyz(mol_name):
    file_path = os.path.join(xyz_folder, mol_name + '.xyz')
    if not os.path.exists(file_path):
        print(f"File not found: {file_path}")  # Debugging line
        return None
    try:
        with open(file_path, 'r') as f:
            xyz_data = f.read()
            return clean_xyz(xyz_data)
    except Exception as e:
        print(f"Error reading {file_path}: {e}")  # Debugging line
        return None

# Function to clean XYZ data (removes unwanted meta information and formats correctly)
def clean_xyz(xyz_str):
    # Remove the quotes if there are any around the whole string
    xyz_str = xyz_str.replace('"', '').strip()  # Remove all quotes and strip whitespace

    # Split the string into lines
    xyz_lines = xyz_str.splitlines()

    # Remove the first two lines (e.g., "10", "./CONFAM_opt.xyz")
    if len(xyz_lines) >= 2:
        xyz_lines = xyz_lines[2:]  # Remove the first two lines

    # Prepare cleaned data by keeping atom names and coordinates only
    cleaned_data = []
    for line in xyz_lines:
        parts = line.split()
        if len(parts) == 4:  # Ensure that there are exactly 4 parts (atom and coordinates)
            cleaned_data.append(" ".join(parts[0:4]))  # Join atom symbol with coordinates

    # Join the cleaned data into a single string
    return "\n".join(cleaned_data)

# Apply the cleaning function to the 'name' column and load the corresponding XYZ data
tqdm.pandas()
df['xyz'] = df['name'].progress_apply(load_xyz)

# Check if xyz column has any data missing or empty
print(f"Number of missing XYZ values: {df['xyz'].isnull().sum()}")

# Save the cleaned dataset to a new CSV file
df.to_csv('formed_xyz_combined.csv', index=False)