tiger.py

import os
import tensorflow as tf
import pandas as pd

GUIDE_LEN = 23
NUCLEOTIDE_TOKENS = dict(zip(['A', 'C', 'G', 'T'], [0, 1, 2, 3]))

# load model
if os.path.exists('model'):
    tiger = tf.keras.models.load_model('model')
else:
    print('no saved model!')
    exit()


def process_data(x):
    x = [item.upper() for item in x]
    number_of_input = len(x) - GUIDE_LEN + 1
    input_gens = []
    for i in range(number_of_input):
        input_gens.append("".join(x[i:i + GUIDE_LEN]))
    merged_token = []
    token_x = [NUCLEOTIDE_TOKENS[item] for item in x]
    for i in range(number_of_input):
        merged_token.extend(token_x[i:i + GUIDE_LEN])
    one_hot_x = tf.one_hot(merged_token, depth=4)
    model_input_x = tf.reshape(one_hot_x, [-1, GUIDE_LEN * 4])
    return input_gens, model_input_x


def tiger_predict(transcript_seq: str):
    # parse transcript sequence into 23-nt target sequences and their one-hot encodings
    target_seq, target_seq_one_hot = process_data(transcript_seq)

    # get predictions
    normalized_lfc = tiger.predict_step(target_seq_one_hot)
    predictions = pd.DataFrame({'Target site': target_seq, 'Normalized LFC': tf.squeeze(normalized_lfc).numpy()})

    return predictions


if __name__ == '__main__':

    # simple test case
    transcript_sequence = 'ACGTACGTACGTACGTACGTACGTACGTACGT'
    df = tiger_predict(transcript_sequence)
    print(df)