Processing Script.py

## Script to sanitize and split HematoxLong2023 dataset

# 1. Load modules 

pip install rdkit
pip install molvs
import pandas as pd
import numpy as np
import urllib.request
import tqdm
import rdkit
from rdkit import Chem
import molvs

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

#2. Download the original dataset

# Download SI Table 2 from
#.  Structural Analysis and Prediction of Hematotoxicity Using Deep Learning Approaches
#.  Teng-Zhi Long, Shao-Hua Shi, Shao Liu, Ai-Ping Lu, Zhao-Qian Liu, Min Li, Ting-Jun Hou*, and Dong-Sheng Cao*
#.  J. Chem. Inf. Model. 2023, 63, 1, 111–125
#.  https://doi.org/10.1021/acs.jcim.2c01088

urllib.request.urlretrieve(
    url = "https://ndownloader.figstatic.com/files/38442996",
    filename = "Long2023_SI2.xls")

Long2023 = pd.read_excel(
    io = "Long2023_SI2.xls",
    sheet_name = "all_data",
    engine = "xlrd",
    names = ['SMILES', 'Y'])

Long2023['ID'] = [f"Long2023_{i}" for i in range(len(Long2023.index))]

#3. Resolve SMILES parse error

# failed to parse smiles, looked it up in ChemSpider and found Octahydro-1H-purine-6-thiol
# (Found by conversion of search term to chemical structure (connectivity match))
# https://www.chemspider.com/Chemical-Structure.114454999.html?rid=8f8bdf21-6482-44e5-adc9-71c5aef41627
Long2023.loc[Long2023['SMILES'] == 'Sc1[nH]cnc-2[n+H]cnc1-2', 'SMILES'] = "SC1NCNC2NCNC21"

# failed to parse smiles, looked it up in Pubchem and found 2-Amino-3,7-dihydropurin-1-ium-6-thione
# https://pubchem.ncbi.nlm.nih.gov/compound/117071895
Long2023.loc[Long2023['SMILES'] == 'Sc1nc(N)nc2[n+H]c[nH]c12', 'SMILES'] = "C1=NC2=C(N1)C(=S)[NH+]=C(N2)N"

# failed to parse smiles, looked it up in ChemSpider and found gentamicin
# (Found by conversion of search term to chemical structure (connectivity match))
# The approved drug gentamicin is a complex of three closely-related aminoglycoside sulfates: C1, C1a and C2
# It's unclear from the smiles, if they used just one of the forms or the mixture of all three
# Clinically, there appears to be a wide range of compositions of C1, C1a, C2 forms (!)
# Here, we'll just use the C1 form (https://www.ebi.ac.uk/chebi/searchId.do;D579C431E9548A30AFBD88A2D05CA777?chebiId=CHEBI:27412)
Long2023.loc[Long2023['SMILES'] == 'O(C1C(O)C(OC2C(O)C([N+H2]C)[C@](O)(C)CO2)C(N)CC1N)C1C([N+H3])CCC(C([N+H2]C)C)O1', 'SMILES'] = \
     "CNC(C)[C@@H]1CC[C@@H](N)[C@H](O1)O[C@@H]1[C@@H](N)C[C@@H](N)[C@H](O[C@H]2OC[C@](C)(O)[C@H](NC)[C@H]2O)[C@H]1O"

# failed to parse smiles, looked it up in ChemSpider and found (2S)-2-Hydroxy-N-isopropyl-3-(1-naphthyloxy)-1-propanaminium
# (Found by conversion of search term to chemical structure (connectivity match))
# https://www.chemspider.com/Chemical-Structure.5296478.html?rid=129380a1-888e-48d0-b240-43646b288842
Long2023.loc[Long2023['SMILES'] == 'O(C[C@@H](O)C[N+H2]C(C)C)c1c2c(ccc1)cccc2', 'SMILES'] = \
     "O[C@@H](C[NH2+]C(C)C)COc2cccc1ccccc12"

# failed to parse smiles, looked it up in ChemSpider, removing charge and H from nitrogens
# matches Periciazine, which I think is what this compound is
# https://www.chemspider.com/Chemical-Structure.4585.html?rid=0694332f-3886-4df1-85d4-8cdb11554c3d
Long2023.loc[Long2023['SMILES'] == 'OC1CC[N+H](CCC[N+@H]2c3c(Sc4c2cccc4)ccc(C#N)c3)CC1', 'SMILES'] = \
     "c1ccc2c(c1)N(c3cc(ccc3S2)C#N)CCCN4CCC(CC4)O"

# failed to parse smiles, looked it up in ChemSpider and found idarubicin
# (Found by conversion of search term to chemical structure (connectivity match))
# Idarubicin, or 4-demethoxydaunorubicin is an anthracycline antileukemic drug.
# https://www.chemspider.com/Chemical-Structure.39117.html?rid=a418032b-d8ab-4200-9226-bcb194221cf0&page_num=0
Long2023.loc[Long2023['SMILES'] == 'O=C(C)[C@@]1(O)CC(OC2OC(C)C(O)C([N+H3])C2)c2c(O)c3C(=O)c4c(C(=O)c3c(O)c2C1)cccc4', 'SMILES'] = \
     "C[C@H]1[C@H]([C@H](C[C@@H](O1)O[C@H]2C[C@@](Cc3c2c(c4c(c3O)C(=O)c5ccccc5C4=O)O)(C(=O)C)O)N)O"

# failed to parse smiles, looked it up in ChemSpider and found (2S)-5-[(6-Methoxy-8-quinolinyl)amino]-2-pentanaminium
# (Found by conversion of search term to chemical structure (full match))
# https://www.chemspider.com/Chemical-Structure.1291555.html?rid=b22e2c45-078e-48d9-8907-22deb37bed0c
Long2023.loc[Long2023['SMILES'] == 'O(C)c1cc(NCCC[C@@H]([N+H3])C)c2ncccc2c1', 'SMILES'] = "C[C@@H](CCCNc1cc(cc2c1nccc2)OC)[NH3+]"

# failed to parse smiles, looked it up in PUbchem
# found diethyl-[2-[[5-[(Z)-(5-fluoro-2-oxo-1H-indol-3-ylidene)methyl]-2,4-dimethyl-1H-pyrrole-3-carbonyl]amino]ethyl]azanium
# https://pubchem.ncbi.nlm.nih.gov/compound/40469377
Long2023.loc[Long2023['SMILES'] == 'Fc1cc2/C(=C/c3c(C)c(C(=O)NCC[N+H](CC)CC)c(C)[nH]3)/C(=O)Nc2cc1', 'SMILES'] = \
   "CC[NH+](CC)CCNC(=O)C1=C(NC(=C1C)C=C2C3=C(C=CC(=C3)F)NC2=O)C"

# failed to parse smiles, looked it up in ChemSpider
# found 1-Methyl-4-{4-[(4-methyl-3-{[4-(3-pyridinyl)-2-pyrimidinyl]amino}phenyl)carbamoyl]benzyl}piperazin-1-ium
# (Found by conversion of search term to chemical structure (full match))
# https://www.chemspider.com/Chemical-Structure.13082883.html?rid=8baa4568-3947-473e-9fdc-8872f411694c#synonymsTab
Long2023.loc[Long2023['SMILES'] == 'O=C([N+H2]c1cc(Nc2nc(-c3cnccc3)ccn2)c(C)cc1)c1ccc(CN2CCN(C)CC2)cc1', 'SMILES'] = \
    "Cc1ccc(cc1Nc2nccc(n2)c3cccnc3)NC(=O)c4ccc(cc4)CN5CC[NH+](CC5)C"

# failed to parse smiles, looked it up in ChemSpider and found 6-Chloro-9-{[(2S)-5-(diethylammonio)-2-pentanyl]amino}-2-methoxyacridinium
# (Found by conversion of search term to chemical structure (full match))
# https://www.chemspider.com/Chemical-Structure.5408834.html?rid=c1cdd9c2-d48b-4cd4-945d-606bf2bff1ee
Long2023.loc[Long2023['SMILES'] == 'Clc1cc2[n+H]c3c(c(N[C@H](CCC[N+H](CC)CC)C)c2cc1)cc(OC)cc3', 'SMILES'] = \
    "Clc3ccc2c(c1cc(OC)ccc1[nH+]c2c3)N[C@@H](C)CCC[NH+](CC)CC"

# failed to parse smiles, looked it up in ChemSpider, removing charge
# found Carvedilol
# (Found by conversion of search term to chemical structure (connectivity match))
# Carvedilol is a beta-blocker medication, that may be prescribed for treating high blood pressure and chronic heart failure with reduced ejection fraction.
# https://www.chemspider.com/Chemical-Structure.2487.html?rid=f8358443-8c32-4e5f-b1d8-47ad13ea8a16&page_num=0
Long2023.loc[Long2023['SMILES'] == 'O(CC[N+H2]C[C@H](O)COc1c-2c([N+H2]c3c-2cccc3)ccc1)c1c(OC)cccc1', 'SMILES'] = \
    "COc1ccccc1OCCNCC(COc2cccc3c2c4ccccc4[nH]3)O"

# failed to parse smiles, looked it up in ChemSpider, found (2S)-1-(2,6-Dimethylphenoxy)-2-propanaminium
# (Found by conversion of search term to chemical structure (full match))
# https://www.chemspider.com/Chemical-Structure.5296573.html?rid=c8c0121e-fad3-4458-8c64-a0baa3e18358
Long2023.loc[Long2023['SMILES'] == 'O(C[C@@H]([N+H3])C)c1c(C)cccc1C', 'SMILES'] = \
    "O(c1c(cccc1C)C)C[C@@H]([NH3+])C"

# failed to parse smiles, looked it up in ChemSpider and found 4-{2-[(5-Cyano-5,5-diphenylpentyl)(methyl)amino]ethyl}-4-methylmorpholin-4-ium
# (Found by conversion of search term to chemical structure (connectivity match))
# https://www.chemspider.com/Chemical-Structure.2310136.html?rid=ce2b48f6-04d6-4df3-ad71-b6507906c7cb
Long2023.loc[Long2023['SMILES'] == '[N+@H](CCCCC(C#N)(c1ccccc1)c1ccccc1)(CC[N+]1(C)CCOCC1)C', 'SMILES'] = \
    "CN(CCCCC(C#N)(c1ccccc1)c2ccccc2)CC[N+]3(CCOCC3)C"

# failed to parse smiles, looked it up in ChemSpider
# found (5E)-3-[(2R,6S)-2,6-Dimethyl-1-piperidinyl]-5-[(4-methoxybenzoyl)imino]-4,5-dihydro-1,2,3-oxadiazol-3-ium-4-ide
# (Found by conversion of search term to chemical structure (full match))
# https://www.chemspider.com/Chemical-Structure.28530638.html?rid=b44dbe3c-24f1-4303-8deb-82b879b7f0e0
Long2023.loc[Long2023['SMILES'] == 'O=C(/N=C1/ON=[N+](N2[C@@H](C)CCC[C@H]2C)[C-H]/1)c1ccc(OC)cc1', 'SMILES'] = \
    "C[C@@H]1CCC[C@@H](N1[n+]2[cH-]/c(=N\C(=O)c3ccc(cc3)OC)/on2)C"

# failed to parse smiles, looked it up in Pubchem
# found ((R)-[(2S)-7-ethenyl-1-azabicyclo[2.2.1]heptan-2-yl]-(6-methoxyquinolin-4-yl)methanol
# https://pubchem.ncbi.nlm.nih.gov/compound/139247817
Long2023.loc[Long2023['SMILES'] == 'O(C)c1cc2c([C@@H](O)[C@@H]3[N+@H]4[C@@H](C=C)[C@H](C3)CC4)ccnc2cc1', 'SMILES'] = \
    "COC1=CC2=C(C=CN=C2C=C1)C(C3CC4CCN3C4C=C)O"

#4. Sanitize with MolVS and print problems

Long2023['X'] = [ \
    rdkit.Chem.MolToSmiles(
        fragment_remover.remove(
        standardizer.standardize(
        rdkit.Chem.MolFromSmiles(
        smiles))))
    for smiles in Long2023['SMILES']]

problems = []
for index, row in tqdm.tqdm(Long2023.iterrows()):
    result = molvs.validate_smiles(row['X'])
    if len(result) == 0:
        continue
    problems.append( (row['ID'], result) )

#   Most are because it includes the salt form and/or it is not neutralized
for id, alert in problems:
    print(f"ID: {id}, problem: {alert[0]}")

#5. Select columns and rename the dataset

Long2023.rename(columns={'X': 'new SMILES', 'Y': 'Label'}, inplace=True)
Long2023[['new SMILES', 'Label']].to_csv('Long2023.csv', index=False)

#6. Import modules to split the dataset

import sys
from rdkit import DataStructs
from rdkit.Chem import AllChem as Chem
from rdkit.Chem import PandasTools   

#7. Split the dataset into test and train

class MolecularFingerprint:
    def __init__(self, fingerprint):
        self.fingerprint = fingerprint

    def __str__(self):
        return self.fingerprint.__str__()

def compute_fingerprint(molecule):
    try:
        fingerprint = Chem.GetMorganFingerprintAsBitVect(molecule, 2, nBits=1024)
        result = np.zeros(len(fingerprint), np.int32)
        DataStructs.ConvertToNumpyArray(fingerprint, result)
        return MolecularFingerprint(result)
    except:
        print("Fingerprints for a structure cannot be calculated")
        return None

def tanimoto_distances_yield(fingerprints, num_fingerprints):
    for i in range(1, num_fingerprints):
        yield [1 - x for x in DataStructs.BulkTanimotoSimilarity(fingerprints[i], fingerprints[:i])]

def cluster_data(fingerprints, num_points, distance_threshold, reordering=False):
    nbr_lists = [None] * num_points
    for i in range(num_points):
        nbr_lists[i] = []

    dist_fun = tanimoto_distances_yield(fingerprints, num_points)
    for i in range(1, num_points):
        dists = next(dist_fun)

        for j in range(i):
            dij = dists[j]
            if dij <= distance_threshold:
                nbr_lists[i].append(j)
                nbr_lists[j].append(i)

    t_lists = [(len(y), x) for x, y in enumerate(nbr_lists)]
    t_lists.sort(reverse=True)

    res = []
    seen = [0] * num_points
    while t_lists:
        _, idx = t_lists.pop(0)
        if seen[idx]:
            continue
        t_res = [idx]
        for nbr in nbr_lists[idx]:
            if not seen[nbr]:
                t_res.append(nbr)
                seen[nbr] = 1
        if reordering:
            nbr_nbr = [nbr_lists[t] for t in t_res]
            nbr_nbr = frozenset().union(*nbr_nbr)
            for x, y in enumerate(t_lists):
                y1 = y[1]
                if seen[y1] or (y1 not in nbr_nbr):
                    continue
                nbr_lists[y1] = set(nbr_lists[y1]).difference(t_res)
                t_lists[x] = (len(nbr_lists[y1]), y1)
            t_lists.sort(reverse=True)
        res.append(tuple(t_res))
    return tuple(res)

def cluster_fingerprints(fingerprints, method="Auto"):
    num_fingerprints = len(fingerprints)

    if method == "Auto":
        method = "TB" if num_fingerprints >= 10000 else "Hierarchy"

    if method == "TB":
        cutoff = 0.56
        print("Butina clustering is selected. Dataset size is:", num_fingerprints)
        clusters = cluster_data(fingerprints, num_fingerprints, cutoff)

    elif method == "Hierarchy":
        import scipy.spatial.distance as ssd
        from scipy.cluster import hierarchy

        print("Hierarchical clustering is selected. Dataset size is:", num_fingerprints)

        av_cluster_size = 8
        dists = []

        for i in range(0, num_fingerprints):
            sims = DataStructs.BulkTanimotoSimilarity(fingerprints[i], fingerprints)
            dists.append([1 - x for x in sims])

        dis_array = ssd.squareform(dists)
        Z = hierarchy.linkage(dis_array)
        average_cluster_size = av_cluster_size
        cluster_amount = int(num_fingerprints / average_cluster_size)
        clusters = hierarchy.cut_tree(Z, n_clusters=cluster_amount)

        clusters = list(clusters.transpose()[0])
        cs = []
        for i in range(max(clusters) + 1):
            cs.append([])

        for i in range(len(clusters)):
            cs[clusters[i]].append(i)
        return cs

def split_dataframe(dataframe, smiles_col_index, fraction_to_train, split_for_exact_fraction=True, cluster_method="Auto"):
    try:
        import math

        smiles_column_name = dataframe.columns[smiles_col_index]
        molecule = 'molecule'
        fingerprint = 'fingerprint'
        group = 'group'
        testing = 'testing'

        try:
            PandasTools.AddMoleculeColumnToFrame(dataframe, smiles_column_name, molecule)
        except:
            print("Exception occurred during molecule generation...")

        dataframe = dataframe.loc[dataframe[molecule].notnull()]
        dataframe[fingerprint] = [compute_fingerprint(m) for m in dataframe[molecule]]
        dataframe = dataframe.loc[dataframe[fingerprint].notnull()]

        fingerprints = [Chem.GetMorganFingerprintAsBitVect(m, 2, nBits=2048) for m in dataframe[molecule]]
        clusters = cluster_fingerprints(fingerprints, method=cluster_method)

        dataframe.drop([molecule, fingerprint], axis=1, inplace=True)

        last_training_index = int(math.ceil(len(dataframe) * fraction_to_train))
        clustered = None
        cluster_no = 0
        mol_count = 0

        for cluster in clusters:
            cluster_no = cluster_no + 1
            try:
                one_cluster = dataframe.iloc[list(cluster)].copy()
            except:
                print("Wrong indexes in Cluster: %i, Molecules: %i" % (cluster_no, len(cluster)))
                continue

            one_cluster.loc[:, 'ClusterNo'] = cluster_no
            one_cluster.loc[:, 'MolCount'] = len(cluster)

            if (mol_count < last_training_index) or (cluster_no < 2):
                one_cluster.loc[:, group] = 'training'
            else:
                one_cluster.loc[:, group] = testing

            mol_count += len(cluster)
            clustered = pd.concat([clustered, one_cluster], ignore_index=True)

        if split_for_exact_fraction:
            print("Adjusting test to train ratio. It may split one cluster")
            clustered.loc[last_training_index + 1:, group] = testing

        print("Clustering finished. Training set size is %i, Test set size is %i, Fraction %.2f" %
              (len(clustered.loc[clustered[group] != testing]),
               len(clustered.loc[clustered[group] == testing]),
               len(clustered.loc[clustered[group] == testing]) / len(clustered)))

    except KeyboardInterrupt:
        print("Clustering interrupted.")

    return clustered


def realistic_split(df, smile_col_index, frac_train, split_for_exact_frac=True, cluster_method = "Auto"):
  return split_dataframe(df.copy(), smile_col_index, frac_train, split_for_exact_frac, cluster_method=cluster_method)

def split_df_into_train_and_test_sets(df):
    df['group'] = df['group'].str.replace(' ', '_')
    df['group'] = df['group'].str.lower()
    train = df[df['group'] == 'training']
    test = df[df['group'] == 'testing']
    return train, test

HematoxLong2023=pd.read_csv("Long2023.csv")
smiles_index = 0  # Because smiles is in the first column
realistic = realistic_split(HematoxLong2023.copy(), smiles_index, 0.75, split_for_exact_frac=True, cluster_method="Auto")
realistic_train, realistic_test = split_df_into_train_and_test_sets(realistic)

#8. Test and train datasets have been made

selected_columns = realistic_train[['new SMILES', 'Label']]
selected_columns.to_csv("HematoxLong2023_train.csv", index=False)
selected_columns = realistic_test[['new SMILES', 'Label']]
selected_columns.to_csv("HematoxLong2023_test.csv", index=False)