#AI夏令营 #Datawhale #夏令营

下载相关库

# 下载相关库
!pip install lightgbm openpyxl rdkit catboost

模块导入

import numpy as np
import pandas as pd
from catboost import CatBoostClassifier, cv, Pool, metrics
from sklearn.model_selection import StratifiedKFold, KFold, GroupKFold
from sklearn.metrics import f1_score
from rdkit import Chem
from rdkit.Chem import Descriptors
from sklearn.feature_extraction.text import TfidfVectorizer
import tqdm, sys, os, gc, re, argparse, warnings
warnings.filterwarnings('ignore')

数据预处理

# 读取训练集和测试集
# 使用 read_excel() 函数从文件中读取训练集数据，文件名为 'traindata-new.xlsx'
train = pd.read_excel('./data/data280993/traindata-new.xlsx')
# 使用 read_excel() 函数从文件中读取测试集数据，文件名为 'testdata-new.xlsx'
test = pd.read_excel('./data/data280993/testdata-new.xlsx')

# test数据不包含 DC50 (nM) 和 Dmax (%)
train = train.drop(['DC50 (nM)', 'Dmax (%)'], axis=1)

# 定义了一个空列表drop_cols，用于存储在测试数据集中非空值小于10个的列名。
drop_cols = []
for f in test.columns:
    if test[f].notnull().sum() < 10:
        drop_cols.append(f)
        
# 使用drop方法从训练集和测试集中删除了这些列，以避免在后续的分析或建模中使用这些包含大量缺失值的列
train = train.drop(drop_cols, axis=1)
test = test.drop(drop_cols, axis=1)

# 使用pd.concat将清洗后的训练集和测试集合并成一个名为data的DataFrame，便于进行统一的特征工程处理
data = pd.concat([train, test], axis=0, ignore_index=True)
cols = data.columns[2:]

data.head()

5 rows × 43 columns

特征工程

# 将SMILES转换为分子对象列表,在转换回去，统一格式
data['smiles_list'] = data['Smiles'].apply(lambda x:[Chem.MolToSmiles(mol, isomericSmiles=True) for mol in [Chem.MolFromSmiles(x)]])
data['smiles_list'] = data['smiles_list'].map(lambda x: ' '.join(x))  

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 704 entries, 0 to 703
Data columns (total 44 columns):
 #   Column                           Non-Null Count  Dtype  
---  ------                           --------------  -----  
 0   uuid                             704 non-null    int64  
 1   Label                            351 non-null    float64
 2   Uniprot                          642 non-null    object 
 3   Target                           704 non-null    object 
 4   E3 ligase                        704 non-null    object 
 5   Name                             277 non-null    object 
 6   Smiles                           704 non-null    object 
 7   Assay (DC50/Dmax)                704 non-null    object 
 8   IC50 (nM, Protac to Target)      153 non-null    float64
 9   Assay (Protac to Target, IC50)   169 non-null    object 
 10  Kd (nM, Protac to Target)        65 non-null     float64
 11  Assay (Protac to Target, Kd)     67 non-null     object 
 12  Assay (Protac to Target, G/H/-T  10 non-null     object 
 13  IC50 (nM, Protac to E3)          32 non-null     float64
 14  Assay (Protac to E3, IC50)       35 non-null     object 
 15  Kd (nM, Protac to E3)            60 non-null     object 
 16  Assay (Protac to E3, Kd)         60 non-null     object 
 17  delta G (kcal/mol, Protac to E3  24 non-null     float64
 18  delta H (kcal/mol, Protac to E3  24 non-null     float64
 19  -T*delta S (kcal/mol, Protac _1  24 non-null     float64
 20  Assay (Protac to E3, G/H/-TS)    24 non-null     object 
 21  Kd (nM, Ternary complex)         27 non-null     object 
 22  Assay (Ternary complex, Kd)      27 non-null     object 
 23  Assay (Ternary complex, G/H/-TS  11 non-null     object 
 24  IC50 (nM, Cellular activities)   131 non-null    float64
 25  Assay (Cellular activities, IC5  190 non-null    object 
 26  EC50 (nM, Cellular activities)   42 non-null     object 
 27  Assay (Cellular activities, EC5  42 non-null     object 
 28  GI50 (nM, Cellular activities)   27 non-null     object 
 29  Assay (Cellular activities, GI5  27 non-null     object 
 30  Article DOI                      704 non-null    object 
 31  Molecular Weight                 704 non-null    float64
 32  Exact Mass                       704 non-null    float64
 33  XLogP3                           704 non-null    float64
 34  Heavy Atom Count                 704 non-null    int64  
 35  Ring Count                       704 non-null    int64  
 36  Hydrogen Bond Acceptor Count     704 non-null    int64  
 37  Hydrogen Bond Donor Count        704 non-null    int64  
 38  Rotatable Bond Count             704 non-null    int64  
 39  Topological Polar Surface Area   704 non-null    float64
 40  Molecular Formula                704 non-null    object 
 41  InChI                            704 non-null    object 
 42  InChI Key                        704 non-null    object 
 43  smiles_list                      704 non-null    object 
dtypes: float64(12), int64(6), object(26)
memory usage: 242.1+ KB

# 使用TfidfVectorizer计算TF-IDF
tfidf = TfidfVectorizer(max_df = 0.9, min_df = 1, sublinear_tf = True)
res = tfidf.fit_transform(data['smiles_list'])

# 将结果转为dataframe格式
tfidf_df = pd.DataFrame(res.toarray())
tfidf_df.columns = [f'smiles_tfidf_{i}' for i in range(tfidf_df.shape[1])]

# 按列合并到data数据
data = pd.concat([data, tfidf_df], axis=1)

data.head()

5 rows × 994 columns

data['Name'].nunique()

208

# 自然数编码
def label_encode(series):
    unique = list(series.unique())
    return series.map(dict(zip(unique, range(series.nunique()))))

for col in cols:
    if data[col].dtype == 'object':
        data[col]  = label_encode(data[col])
        
train = data[data.Label.notnull()].reset_index(drop=True)
test = data[data.Label.isnull()].reset_index(drop=True)

# 特征筛选
features = [f for f in train.columns if f not in ['uuid','Label','smiles_list']]

# 构建训练集和测试集
x_train = train[features]
x_test = test[features]

# 训练集标签
y_train = train['Label'].astype(int)

x_train.isna().sum()[x_train.isna().sum()>0]

IC50 (nM, Protac to Target)        283
Kd (nM, Protac to Target)          329
IC50 (nM, Protac to E3)            335
delta G (kcal/mol, Protac to E3    351
delta H (kcal/mol, Protac to E3    351
-T*delta S (kcal/mol, Protac _1    351
IC50 (nM, Cellular activities)     281
dtype: int64

x_train["Kd (nM, Protac to Target)"]

0        NaN
1        NaN
2        NaN
3        NaN
4      375.0
       ...  
346      NaN
347      NaN
348      NaN
349      NaN
350      NaN
Name: Kd (nM, Protac to Target), Length: 351, dtype: float64

训练、预测

def cv_model(clf, train_x, train_y, test_x, clf_name, seed=2022):
    
    kf = KFold(n_splits=5, shuffle=True, random_state=seed)

    train = np.zeros(train_x.shape[0])
    test = np.zeros(test_x.shape[0])

    cv_scores = []

    for i, (train_index, valid_index) in enumerate(kf.split(train_x, train_y)):
        print('************************************ {} {}************************************'.format(str(i+1), str(seed)))
        trn_x, trn_y, val_x, val_y = train_x.iloc[train_index], train_y[train_index], train_x.iloc[valid_index], train_y[valid_index]
               
        params = {'learning_rate': 0.1, 'depth': 6, 'l2_leaf_reg': 10, 'bootstrap_type':'Bernoulli','random_seed':seed,
                  'od_type': 'Iter', 'od_wait': 100, 'random_seed': 11, 'allow_writing_files': False, 'task_type':'CPU'}

        model = clf(iterations=20000, **params, eval_metric='AUC')
        model.fit(trn_x, trn_y, eval_set=(val_x, val_y),
                  metric_period=100,
                  cat_features=[], 
                  use_best_model=True, 
                  verbose=1)

        val_pred  = model.predict_proba(val_x)[:,1]
        test_pred = model.predict_proba(test_x)[:,1]
            
        train[valid_index] = val_pred
        test += test_pred / kf.n_splits
        cv_scores.append(f1_score(val_y, np.where(val_pred>0.5, 1, 0)))
        
        print(cv_scores)
       
    print("%s_score_list:" % clf_name, cv_scores)
    print("%s_score_mean:" % clf_name, np.mean(cv_scores))
    print("%s_score_std:" % clf_name, np.std(cv_scores))
    return train, test
    
cat_train, cat_test = cv_model(CatBoostClassifier, x_train, y_train, x_test, "cat")

************************************ 1 2022************************************
0:	test: 0.7420034	best: 0.7420034 (0)	total: 53.8ms	remaining: 17m 55s


Warning: Overfitting detector is active, thus evaluation metric is calculated on every iteration. 'metric_period' is ignored for evaluation metric.


100:	test: 0.9082492	best: 0.9116162 (58)	total: 557ms	remaining: 1m 49s
200:	test: 0.9082492	best: 0.9158249 (139)	total: 1.09s	remaining: 1m 47s
Stopped by overfitting detector  (100 iterations wait)

bestTest = 0.9158249158
bestIteration = 139

Shrink model to first 140 iterations.
[0.9052631578947369]
************************************ 2 2022************************************


Warning: Overfitting detector is active, thus evaluation metric is calculated on every iteration. 'metric_period' is ignored for evaluation metric.


0:	test: 0.7957428	best: 0.7957428 (0)	total: 6.33ms	remaining: 2m 6s
100:	test: 0.9248188	best: 0.9248188 (100)	total: 599ms	remaining: 1m 57s
200:	test: 0.9483696	best: 0.9510870 (167)	total: 1.2s	remaining: 1m 58s
300:	test: 0.9547101	best: 0.9547101 (265)	total: 1.91s	remaining: 2m 5s
400:	test: 0.9565217	best: 0.9574275 (324)	total: 2.6s	remaining: 2m 7s
Stopped by overfitting detector  (100 iterations wait)

bestTest = 0.9574275362
bestIteration = 324

Shrink model to first 325 iterations.
[0.9052631578947369, 0.9072164948453608]
************************************ 3 2022************************************
0:	test: 0.7631822	best: 0.7631822 (0)	total: 5.48ms	remaining: 1m 49s


Warning: Overfitting detector is active, thus evaluation metric is calculated on every iteration. 'metric_period' is ignored for evaluation metric.


100:	test: 0.8940796	best: 0.9033302 (65)	total: 601ms	remaining: 1m 58s
Stopped by overfitting detector  (100 iterations wait)

bestTest = 0.9033302498
bestIteration = 65

Shrink model to first 66 iterations.
[0.9052631578947369, 0.9072164948453608, 0.88]
************************************ 4 2022************************************


Warning: Overfitting detector is active, thus evaluation metric is calculated on every iteration. 'metric_period' is ignored for evaluation metric.


0:	test: 0.8167347	best: 0.8167347 (0)	total: 6.05ms	remaining: 2m
100:	test: 0.8897959	best: 0.8946939 (87)	total: 673ms	remaining: 2m 12s
200:	test: 0.8930612	best: 0.8971429 (155)	total: 1.28s	remaining: 2m 6s
300:	test: 0.8963265	best: 0.8987755 (225)	total: 1.81s	remaining: 1m 58s
Stopped by overfitting detector  (100 iterations wait)

bestTest = 0.8987755102
bestIteration = 225

Shrink model to first 226 iterations.
[0.9052631578947369, 0.9072164948453608, 0.88, 0.8571428571428571]
************************************ 5 2022************************************
0:	test: 0.8423913	best: 0.8423913 (0)	total: 5.73ms	remaining: 1m 54s


Warning: Overfitting detector is active, thus evaluation metric is calculated on every iteration. 'metric_period' is ignored for evaluation metric.


100:	test: 0.9547101	best: 0.9655797 (46)	total: 670ms	remaining: 2m 12s
Stopped by overfitting detector  (100 iterations wait)

bestTest = 0.9655797101
bestIteration = 46

Shrink model to first 47 iterations.
[0.9052631578947369, 0.9072164948453608, 0.88, 0.8571428571428571, 0.9052631578947369]
cat_score_list: [0.9052631578947369, 0.9072164948453608, 0.88, 0.8571428571428571, 0.9052631578947369]
cat_score_mean: 0.8909771335555383
cat_score_std: 0.019683259686381092

#pd.DataFrame(
#    {
#        'uuid': test['uuid'],
#        'Label': np.where(cat_test>0.5, 1, 0)
#    }
#).to_csv('submit1.csv', index=None)