tiger.py

import os
import gzip
import numpy as np
import pandas as pd
import tensorflow as tf
from Bio import SeqIO

GUIDE_LEN = 23
CONTEXT_5P = 3
CONTEXT_3P = 0
TARGET_LEN = CONTEXT_5P + GUIDE_LEN + CONTEXT_3P
NUCLEOTIDE_TOKENS = dict(zip(['A', 'C', 'G', 'T'], [0, 1, 2, 3]))
NUCLEOTIDE_COMPLEMENT = dict(zip(['A', 'C', 'G', 'T'], ['T', 'G', 'C', 'A']))
NUM_TOP_GUIDES = 10
NUM_MISMATCHES = 3


def sequence_complement(sequence: list):
    return [''.join([NUCLEOTIDE_COMPLEMENT[nt] for nt in list(seq)]) for seq in sequence]


def one_hot_encode_sequence(sequence: list, add_context_padding: bool = False):

    # stack list of sequences into a tensor
    sequence = tf.ragged.stack([tf.constant(list(seq)) for seq in sequence], axis=0)

    # tokenize sequence
    nucleotide_table = tf.lookup.StaticVocabularyTable(
        initializer=tf.lookup.KeyValueTensorInitializer(
            keys=tf.constant(list(NUCLEOTIDE_TOKENS.keys()), dtype=tf.string),
            values=tf.constant(list(NUCLEOTIDE_TOKENS.values()), dtype=tf.int64)),
        num_oov_buckets=1)
    sequence = tf.RaggedTensor.from_row_splits(values=nucleotide_table.lookup(sequence.values),
                                               row_splits=sequence.row_splits).to_tensor(255)

    # add context padding if requested
    if add_context_padding:
        pad_5p = 255 * tf.ones([sequence.shape[0], CONTEXT_5P], dtype=sequence.dtype)
        pad_3p = 255 * tf.ones([sequence.shape[0], CONTEXT_3P], dtype=sequence.dtype)
        sequence = tf.concat([pad_5p, sequence, pad_3p], axis=1)

    # one-hot encode
    sequence = tf.one_hot(sequence, depth=4)

    return sequence


def process_data(transcript_seq: str):

    # convert to upper case
    transcript_seq = transcript_seq.upper()

    # get all target sites
    target_seq = [transcript_seq[i: i + TARGET_LEN] for i in range(len(transcript_seq) - TARGET_LEN)]

    # prepare guide sequences
    guide_seq = sequence_complement([seq[CONTEXT_5P:len(seq) - CONTEXT_3P] for seq in target_seq])

    # model inputs
    model_inputs = tf.concat([
        tf.reshape(one_hot_encode_sequence(target_seq, add_context_padding=False), [len(target_seq), -1]),
        tf.reshape(one_hot_encode_sequence(guide_seq, add_context_padding=True), [len(guide_seq), -1]),
        ], axis=-1)

    return target_seq, guide_seq, model_inputs


def tiger_predict(transcript_seq: str):

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

    # parse transcript sequence
    target_seq, guide_seq, model_inputs = process_data(transcript_seq)

    # get predictions
    normalized_lfc = tiger.predict_step(model_inputs)
    predictions = pd.DataFrame({'Guide': guide_seq, 'Normalized LFC': tf.squeeze(normalized_lfc).numpy()})
    predictions = predictions.set_index('Guide').sort_values('Normalized LFC')

    return predictions


def find_off_targets(guides, batch_size=1000):
    with gzip.open(os.path.join('transcripts', 'gencode.v19.pc_transcripts.fa.gz'), 'rt') as file:
        df_transcripts = pd.DataFrame([(t.id, str(t.seq)) for t in SeqIO.parse(file, 'fasta')], columns=['id', 'seq'])
    df_transcripts['id'] = df_transcripts['id'].apply(lambda s: s.split('|')[4])
    df_transcripts.set_index('id', inplace=True)

    # one-hot encode guides to form a filter
    guide_filter = one_hot_encode_sequence(sequence_complement(guides), add_context_padding=False)
    guide_filter = tf.transpose(guide_filter, [1, 2, 0])

    # loop over transcripts in batches
    i = 0
    print('Scanning for off-targets')
    df_off_targets = pd.DataFrame()
    while i < len(df_transcripts):
        # select batch
        df_batch = df_transcripts.iloc[i:min(i + batch_size, len(df_transcripts))]
        i += batch_size

        # find and log off-targets
        transcripts = one_hot_encode_sequence(df_batch['seq'].values.tolist(), add_context_padding=False)
        num_mismatches = GUIDE_LEN - tf.nn.conv1d(transcripts, guide_filter, stride=1, padding='SAME')
        loc_off_targets = tf.where(num_mismatches <= NUM_MISMATCHES).numpy()
        df_off_targets = pd.concat([df_off_targets, pd.DataFrame({
            'Guide': np.array(guides)[loc_off_targets[:, 2]],
            'Isoform': df_batch.index.values[loc_off_targets[:, 0]],
            'Mismatches': tf.gather_nd(num_mismatches, loc_off_targets).numpy().astype(int),
            'Midpoint': loc_off_targets[:, 1],
            'Target': df_batch['seq'].values[loc_off_targets[:, 0]],
        })])

        # progress update
        print('\rPercent complete: {:.2f}%'.format(100 * min(i / len(df_transcripts), 1)), end='')
    print('')

    # trim transcripts to targets
    dict_off_targets = df_off_targets.to_dict('records')
    for row in dict_off_targets:
        start_location = row['Midpoint'] - (GUIDE_LEN // 2) - CONTEXT_5P
        row['Target'] = row['Target'][start_location:start_location + TARGET_LEN]
        if row['Mismatches'] == 0:
            assert row['Guide'] == sequence_complement([row['Target'][CONTEXT_5P:TARGET_LEN-CONTEXT_3P]])[0]
    df_off_targets = pd.DataFrame(dict_off_targets)

    return df_off_targets


if __name__ == '__main__':

    # simple test case
    transcript_sequence = 'ATGCAGGACGCGGAGAACGTGGCGGTGCCCGAGGCGGCCGAGGAGCGCGC'.lower()  # first 50 from EIF3B-003's CDS
    sorted_predictions = tiger_predict(transcript_sequence)

    # report top guides only
    sorted_predictions = sorted_predictions.iloc[:NUM_TOP_GUIDES]
    print(sorted_predictions)

    # scan for off-targets for top guides
    off_targets = find_off_targets(sorted_predictions.index.values.tolist())
    print(off_targets)