harshilshrivastava / dataset-1

Software Requirements

python 3

python packages: pandas, numpy, math, keras, sklearn, matplotlib

jupyter notebook

Datasets

link:

Steps to follow:

The project goal is to develop a framework to predict if a biological sequence belongs to a promoter or non-promoter category. Steps are as follows:

1. Remove duplicate sequences from promoter data.
2. Generate shuffle data from promoter sequences.
3. Perform k-merization.
4. Frequency based tokenization
5. Building and training classification algorithms:

RandomForestClassifier (from sklearn.ensemble)

LSTM (from keras.layer)

CNN (from keras.layers.convolutional)

1. Performance evaluation

Code snippets to develop a framework

Drop Duplicates:

X_drop_dup = X.drop_duplicates()

idx = X_drop_dup.index

data = np.array(df)

df= pd.DataFrame(data[idx])

Shuffle Data Snippet:

def shuffler(sequence, k_let):

length = [sequence[i:i+k_let] for i in range(0,len(sequence),k_let)]

np.random.shuffle(length)

return ''.join(length)

K-merization:

k=2 #K-mer size 2, 4 or8

def getKmers(X, size=k):

return [X[x:x+size].lower() for x in range(len(X) - size + 1)]

CNN architecture:

model = Sequential()

model.add(Embedding(vocab_size,128,input_length=max_len))

model.add(Conv1D(filters=128, kernel_size=5,padding='same'))

model.add(MaxPooling1D(pool_size=4))

model.add(Conv1D(filters=64, kernel_size=5, padding='same'))

model.add(MaxPooling1D(pool_size=4))

model.add(Conv1D(filters=32, kernel_size=5, padding='same'))

model.add(MaxPooling1D(pool_size=4))

model.add(Dense(1024, activation='relu'))

model.add(Dropout(0.2))

model.add(Dense(512, activation='relu'))

model.add(Dropout(0.2))

model.add(Dense(128, activation='relu'))

model.add(Dropout(0.2))

model.add(Flatten())

model.add(Dense(1, activation='sigmoid'))

LSTM architecture:

model = Sequential()

model.add(Embedding(vocab_size, 50 ,input_length=max_len))

model.add(LSTM(128))

model.add(Dense(64, activation='relu'))

model.add(Dropout(0.5))

model.add(Dense(1, activation='sigmoid'))

For binary classification:

Loss function: binary_crossentropy,

Optimizer: adam

number of epochs: 10

For multispecies classification:

Loss function: sparse_categorical_crossentropy

Optimizer: adam

number of epochs: 10