CODEWITHSUNDEEP
pythonarraysscikit-learnpython-itertoolsone-hot-encoding
Xiong89
Xiong89

Reputation: 777

How to generate one hot encoding for DNA sequences?

I would like to generate one hot encoding for a set of DNA sequences. For example the sequence ACGTCCA can be represented as below in a transpose manner. But the code below will generate the one hot encoding in horizontal way in which I would prefer it in vertical form. Can anyone help me?

ACGTCCA 
1000001 - A
0100110 - C 
0010000 - G
0001000 - T

Example code:

from sklearn.preprocessing import OneHotEncoder
import itertools

# two example sequences
seqs = ["ACGTCCA","CGGATTG"]


# split sequences to tokens
tokens_seqs = [seq.split("\\") for seq in seqs]

# convert list of of token-lists to one flat list of tokens
# and then create a dictionary that maps word to id of word,
# like {A: 1, B: 2} here
all_tokens = itertools.chain.from_iterable(tokens_seqs)
word_to_id = {token: idx for idx, token in enumerate(set(all_tokens))}

# convert token lists to token-id lists, e.g. [[1, 2], [2, 2]] here
token_ids = [[word_to_id[token] for token in tokens_seq] for tokens_seq in tokens_seqs]

# convert list of token-id lists to one-hot representation
vec = OneHotEncoder(n_values=len(word_to_id))
X = vec.fit_transform(token_ids)

print X.toarray()

However, the code gives me output:

[[ 0.  1.]
 [ 1.  0.]]

Expected output:

[[1. 0. 0. 0. 0. 0. 1. 0. 1. 0. 0. 1. 1. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0.]
[0. 0. 0. 1. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 1. 1. 0. 0. 0. 1. 0. 0. 0. 0. 1. 1. 0.]]

Upvotes: 10

Views: 7700

Answers (3)

Rajan saha Raju
Rajan saha Raju

Reputation: 840

from keras.utils import to_categorical

def one_hot_encoding(seq):
    mp = dict(zip('ACGT', range(4)))
    seq_2_number = [mp[nucleotide] for nucleotide in seq]
    return to_categorical(seq_2_number, num_classes=4, dtype='int32').flatten()

Upvotes: 0

DrIDK
DrIDK

Reputation: 7973

def one_hot_encode(seq):
    mapping = dict(zip("ACGT", range(4)))    
    seq2 = [mapping[i] for i in seq]
    return np.eye(4)[seq2]

one_hot_encode("AACGT")

## Output: 
array([[1., 0., 0., 0.],
   [1., 0., 0., 0.],
   [0., 1., 0., 0.],
   [0., 0., 1., 0.],
   [0., 0., 0., 1.]])

Upvotes: 17

John Zwinck
John Zwinck

Reputation: 249582

I suggest doing it a slightly more manual way:

import numpy as np

seqs = ["ACGTCCA","CGGATTG"]

CHARS = 'ACGT'
CHARS_COUNT = len(CHARS)

maxlen = max(map(len, seqs))
res = np.zeros((len(seqs), CHARS_COUNT * maxlen), dtype=np.uint8)

for si, seq in enumerate(seqs):
    seqlen = len(seq)
    arr = np.chararray((seqlen,), buffer=seq)
    for ii, char in enumerate(CHARS):
        res[si][ii*seqlen:(ii+1)*seqlen][arr == char] = 1

print res

This gives you your desired result:

[[1 0 0 0 0 0 1 0 1 0 0 1 1 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0]
 [0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 1 1 0 0 0 1 0 0 0 0 1 1 0]]

Upvotes: 6

Related Questions