Predicting peptide-protein binding with a random forest

from tdc.benchmark_group.protein_peptide_group import ProteinPeptideGroup
group = ProteinPeptideGroup()
train, val = group.get_train_valid_split() # val dataset will be empty. use the train dataset if fine-tuning desired.
test = group.get_test()
X_train, y_train = train[0], train[1]
X_test, y_test = test[0], test[1]

X_train.head()

y_train.value_counts()

import protpy as protpy
import pandas as pd

function_list = [protpy.amino_acid_composition, 
                 protpy.dipeptide_composition,
                 protpy.tripeptide_composition,
                 protpy.moreaubroto_autocorrelation,
                 #using default parameters: lag=30, properties=["CIDH920105", "BHAR880101", "CHAM820101", "CHAM820102", "CHOC760101", "BIGC670101", "CHAM810101", "DAYM780201"], normalize=True
                 protpy.moran_autocorrelation,
                 #using default parameters: lag=30, properties=["CIDH920105", "BHAR880101", "CHAM820101", "CHAM820102", "CHOC760101", "BIGC670101", "CHAM810101", "DAYM780201"], normalize=True
                 protpy.geary_autocorrelation,
                 #using default parameters: lag=30, properties=["CIDH920105", "BHAR880101", "CHAM820101", "CHAM820102", "CHOC760101", "BIGC670101", "CHAM810101", "DAYM780201"], normalize=True,
                 protpy.conjoint_triad,
                 #protpy.sequence_order_coupling_number_, ### <--------- this returns a float, i will have to handle it separately 
                 protpy.sequence_order_coupling_number,
                 #using default parameters: lag=30, distance_matrix="schneider-wrede"
                 protpy.quasi_sequence_order,
                 #using default parameters: lag=30, weight=0.1, distance_matrix="schneider-wrede"
                ]

X_train = X_train[['X1','Protein Target_ID']]
X_train = X_train.reset_index()

for function in function_list:
    df_list=[]
    for index, sequence in enumerate(X_train['X1']):
        AA = function(sequence)
        df_list.append(AA)
    tempdf = pd.concat(df_list)
    tempdf=tempdf.reset_index(drop=True)
    X_train = X_train.join(tempdf)
X_train['SOCN'] = X_train['X1'].apply(protpy.sequence_order_coupling_number_)

#There's probably a faster way to do this, but this way works - I had trouble applying protpy functions to entire dataframes in parallel. 

X_train.head()

y_train =  y_train.reset_index(drop=True)
y_train[y_train=='1']

X_train.iloc[[155, 391, 412]]

MDM2_list = X_train.index[X_train['Protein Target_ID']=='MDM2'].tolist()
X_train_MDM2 = X_train[X_train['Protein Target_ID']=='MDM2']
y_trian_MDM2 = y_train.iloc[MDM2_list]
#Also drop the Protein Target ID column because... we don't need it anymore.
X_train_MDM2 = X_train_MDM2.drop(columns=['index','X1','Protein Target_ID'])
X_train_MDM2.head()

import imblearn.over_sampling
sm = imblearn.over_sampling.SMOTE(random_state=42, k_neighbors=2)
X_res, y_res = sm.fit_resample(X_train_MDM2, y_trian_MDM2)
X_res

import numpy as np
#from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score

rf_classifier_smote = RandomForestClassifier(random_state=42)
#X_train, y_train = X_res, y_res
rf_classifier_smote.fit(X_res, y_res)

#function_list = [protpy.amino_acid_composition, 
#                 protpy.dipeptide_composition,
#                 protpy.tripeptide_composition,
#                 protpy.moreaubroto_autocorrelation,
#                 #using default parameters: lag=30, properties=["CIDH920105", "BHAR880101", "CHAM820101", "CHAM820102", "CHOC760101", "BIGC670101", "CHAM810101", "DAYM780201"], normalize=True
#                 protpy.moran_autocorrelation,
#                 #using default parameters: lag=30, properties=["CIDH920105", "BHAR880101", "CHAM820101", "CHAM820102", "CHOC760101", "BIGC670101", "CHAM810101", "DAYM780201"], normalize=True
#                 protpy.geary_autocorrelation,
#                 #using default parameters: lag=30, properties=["CIDH920105", "BHAR880101", "CHAM820101", "CHAM820102", "CHOC760101", "BIGC670101", "CHAM810101", "DAYM780201"], normalize=True,
#                 protpy.conjoint_triad,
#                 #protpy.sequence_order_coupling_number_, ### <--------- this returns a float, i will have to handle it separately 
#                 protpy.sequence_order_coupling_number,
#                 #using default parameters: lag=30, distance_matrix="schneider-wrede"
#                 protpy.quasi_sequence_order,
#                 #using default parameters: lag=30, weight=0.1, distance_matrix="schneider-wrede"
#                ]
#
X_test = X_test[['X1','Protein Target_ID']]
X_test = X_test.reset_index()

for function in function_list:
    df_list=[]
    for index, sequence in enumerate(X_test['X1']):
        AA = function(sequence)
        df_list.append(AA)
    tempdf = pd.concat(df_list)
    tempdf=tempdf.reset_index(drop=True)
    X_test = X_test.join(tempdf)
X_test['SOCN'] = X_test['X1'].apply(protpy.sequence_order_coupling_number_)

#There's probably a faster way to do this, but this way works - I had trouble applying protpy functions to entire dataframes in parallel. 

X_test.head()

X_test = X_test.drop(columns=['index','X1','Protein Target_ID'])

predictions = rf_classifier_smote.predict(X_test)
## The evaluate function requires a very particular structure for the data ##
predictions = predictions.astype(str)
predictions = predictions.astype(object)
out = group.evaluate(predictions)
out