src/utils.py

#!/bin/python

import pandas as pd
import numpy as np
from rdkit import Chem
from rdkit.Chem import Descriptors

# this function is to fill rows in which the protein structure is the same, but there are multiple ligands (appearing as multiple rows) bound to the protein
def fill_nans_iteratively(df):
    """
    Iteratively fill NaN values in rows that have NaNs in all columns except
    'Ligand ID', 'Ligand Name', 'Ligand SMILES', and 'Protein'
    
    Parameters:
    df: pandas DataFrame
    
    Returns:
    DataFrame with filled values
    """
    
    # make a copy to avoid modifying the original
    df_filled = df.copy()
    
    # define columns to exclude from the NaN check
    exclude_cols = ['Ligand ID', 'Ligand Name', 'Ligand SMILES', 'Protein']
    
    # get all other columns
    other_cols = [col for col in df_filled.columns if col not in exclude_cols]
    
    # iterate through rows starting from index 1 (second row)
    for i in range(1, len(df_filled)):
        
        # check if current row has NaN in all other columns
        current_row_other_cols = df_filled.loc[df_filled.index[i], other_cols]
        
        if current_row_other_cols.isna().all():
            
            # get the previous row's values for the other columns
            prev_row_values = df_filled.loc[df_filled.index[i-1], other_cols]
            
            # fill NaN values in current row with previous row's values
            df_filled.loc[df_filled.index[i], other_cols] = prev_row_values
            
    return df_filled


def calculate_mol_weight(smiles):
    """Helper function to calculate molecular weight from SMILES"""
    if pd.isna(smiles) or not isinstance(smiles, str):
        return np.nan
    try:
        mol = Chem.MolFromSmiles(smiles)
        if mol is not None:
            return Descriptors.MolWt(mol)
        else:
            return np.nan
    except:
        return np.nan

# this function filters by molecular weight to remove rows where the ligands are crystallization agents, ions, cofactors
def filter_by_molecular_weight(df, min_weight=150, max_weight=1000):
    """
    Filter df by ligands within molecular weight range
    """
    df['Molecular Weight'] = df['Ligand SMILES'].apply(calculate_mol_weight)
    
    successful = df['Molecular Weight'].notna().sum()
    total = len(df)
    
    if successful == 0:
        print("WARNING: No molecules were successfully processed!")
        # show some examples of the SMILES strings for debugging
        print("Sample SMILES strings:")
        print(df['Ligand Smiles'].head(10).tolist())
        return pd.DataFrame()
    
    mw_filtered_df = df[
        (df['Molecular Weight'].notna()) & 
        (df['Molecular Weight'] >= min_weight) & 
        (df['Molecular Weight'] <= max_weight)
    ]
    
    # finally, remove any where pdb id is not present
    mw_filtered_df = mw_filtered_df[(mw_filtered_df['Entry ID'].notna())]
    
    return mw_filtered_df