In [2]:

    

import rdkit
import pandas as pd
import numpy as np
from scipy.stats.stats import pearsonr 
from sklearn import linear_model
from heapq import nlargest
from rdkit import Chem


    

In [5]:

    

fileName = "train.csv"
train = pd.read_csv(fileName)


    

In [6]:

    

len(train)


    

    
Out[6]:





1000000

In [3]:

    

train = train[start:]
print len(train.smiles)


    

    




50000

In [4]:

    

from rdkit.Chem import AllChem
m = Chem.MolFromSmiles(train.smiles[start])
x1 = AllChem.GetMorganFingerprintAsBitVect(m, 1, nBits=512, useFeatures=True)
x2= AllChem.GetMorganFingerprintAsBitVect(m, 2, nBits=512, useFeatures=True)
x3 = AllChem.GetMorganFingerprintAsBitVect(m, 3, nBits=1024, useFeatures=True)
x4 = AllChem.GetHashedAtomPairFingerprintAsBitVect(m, nBits=256)
x5 = AllChem.GetHashedTopologicalTorsionFingerprintAsBitVect(m , nBits=256)
new_row = x1 + x2 + x3 + x4 + x5
features = new_row


    

In [5]:

    

for i in range(start+1,end):
    if(i %10000 == 0):
        print i
    m = Chem.MolFromSmiles(train.smiles[i])
    x1 = AllChem.GetMorganFingerprintAsBitVect(m, 1, nBits=512, useFeatures=True)
    x2= AllChem.GetMorganFingerprintAsBitVect(m, 2, nBits=512, useFeatures=True)
    x3 = AllChem.GetMorganFingerprintAsBitVect(m, 3, nBits=1024, useFeatures=True)
    x4 = AllChem.GetHashedAtomPairFingerprintAsBitVect(m, nBits=256)
    x5 = AllChem.GetHashedTopologicalTorsionFingerprintAsBitVect(m , nBits=256)
    new_row = x1 + x2 + x3 + x4 + x5
    features = features + new_row


    

    




610000
620000
630000
640000

In [6]:

    

len(features)


    

    
Out[6]:





128000000

In [7]:

    

new_feats = np.array(features)


    

In [8]:

    

new_feats = new_feats.reshape(end-start,2560)


    

In [9]:

    

new_feats.shape


    

    
Out[9]:





(50000, 2560)

In [10]:

    

a = pd.DataFrame(new_feats)


    

In [11]:

    

a.to_csv("train"  + str(start))


    

In [ ]: