Create plant-multi-species-genomes.py
Browse files- plant-multi-species-genomes.py +161 -0
plant-multi-species-genomes.py
ADDED
@@ -0,0 +1,161 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
"""Script for the plant multi-species genomes dataset. This dataset contains the genomes
|
3 |
+
from 48 different species."""
|
4 |
+
|
5 |
+
from typing import List
|
6 |
+
import datasets
|
7 |
+
import pandas as pd
|
8 |
+
from Bio import SeqIO
|
9 |
+
|
10 |
+
|
11 |
+
# Find for instance the citation on arxiv or on the dataset repo/website
|
12 |
+
_CITATION = """"""
|
13 |
+
|
14 |
+
# You can copy an official description
|
15 |
+
_DESCRIPTION = """\
|
16 |
+
Dataset made of diverse genomes available on NCBI and coming from 48 different species.
|
17 |
+
Test and validation are made of 2 species each. The rest of the genomes are used for training.
|
18 |
+
Default configuration "6kbp" yields chunks of 6.2kbp (100bp overlap on each side).The chunks of DNA are cleaned and processed so that
|
19 |
+
they can only contain the letters A, T, C, G and N.
|
20 |
+
"""
|
21 |
+
|
22 |
+
_HOMEPAGE = "" #"https://www.ncbi.nlm.nih.gov/"
|
23 |
+
|
24 |
+
_LICENSE = "" #"https://www.ncbi.nlm.nih.gov/home/about/policies/"
|
25 |
+
|
26 |
+
_CHUNK_LENGTHS = [6000,]
|
27 |
+
|
28 |
+
|
29 |
+
def filter_fn(char: str) -> str:
|
30 |
+
"""
|
31 |
+
Transforms any letter different from a base nucleotide into an 'N'.
|
32 |
+
"""
|
33 |
+
if char in {'A', 'T', 'C', 'G'}:
|
34 |
+
return char
|
35 |
+
else:
|
36 |
+
return 'N'
|
37 |
+
|
38 |
+
|
39 |
+
def clean_sequence(seq: str) -> str:
|
40 |
+
"""
|
41 |
+
Process a chunk of DNA to have all letters in upper and restricted to
|
42 |
+
A, T, C, G and N.
|
43 |
+
"""
|
44 |
+
seq = seq.upper()
|
45 |
+
seq = map(filter_fn, seq)
|
46 |
+
seq = ''.join(list(seq))
|
47 |
+
return seq
|
48 |
+
|
49 |
+
|
50 |
+
class PlantMultiSpeciesGenomesConfig(datasets.BuilderConfig):
|
51 |
+
"""BuilderConfig for the Plant Multi Species Pre-training Dataset."""
|
52 |
+
|
53 |
+
def __init__(self, *args, chunk_length: int, overlap: int = 100, **kwargs):
|
54 |
+
"""BuilderConfig for the multi species genomes.
|
55 |
+
Args:
|
56 |
+
chunk_length (:obj:`int`): Chunk length.
|
57 |
+
overlap: (:obj:`int`): Overlap in base pairs for two consecutive chunks (defaults to 100).
|
58 |
+
**kwargs: keyword arguments forwarded to super.
|
59 |
+
"""
|
60 |
+
num_kbp = int(chunk_length/1000)
|
61 |
+
super().__init__(
|
62 |
+
*args,
|
63 |
+
name=f'{num_kbp}kbp',
|
64 |
+
**kwargs,
|
65 |
+
)
|
66 |
+
self.chunk_length = chunk_length
|
67 |
+
self.overlap = overlap
|
68 |
+
|
69 |
+
|
70 |
+
class PlantMultiSpeciesGenomes(datasets.GeneratorBasedBuilder):
|
71 |
+
"""Genomes from 48 species, filtered and split into chunks of consecutive
|
72 |
+
nucleotides. 2 genomes are taken for test, 2 for validation and 44
|
73 |
+
for training."""
|
74 |
+
|
75 |
+
VERSION = datasets.Version("1.1.0")
|
76 |
+
BUILDER_CONFIG_CLASS = PlantMultiSpeciesGenomesConfig
|
77 |
+
BUILDER_CONFIGS = [PlantMultiSpeciesGenomesConfig(chunk_length=chunk_length) for chunk_length in _CHUNK_LENGTHS]
|
78 |
+
DEFAULT_CONFIG_NAME = "6kbp"
|
79 |
+
|
80 |
+
def _info(self):
|
81 |
+
|
82 |
+
features = datasets.Features(
|
83 |
+
{
|
84 |
+
"sequence": datasets.Value("string"),
|
85 |
+
"description": datasets.Value("string"),
|
86 |
+
"start_pos": datasets.Value("int32"),
|
87 |
+
"end_pos": datasets.Value("int32"),
|
88 |
+
}
|
89 |
+
)
|
90 |
+
return datasets.DatasetInfo(
|
91 |
+
# This is the description that will appear on the datasets page.
|
92 |
+
description=_DESCRIPTION,
|
93 |
+
# This defines the different columns of the dataset and their types
|
94 |
+
features=features,
|
95 |
+
# Homepage of the dataset for documentation
|
96 |
+
homepage=_HOMEPAGE,
|
97 |
+
# License for the dataset if available
|
98 |
+
license=_LICENSE,
|
99 |
+
# Citation for the dataset
|
100 |
+
citation=_CITATION,
|
101 |
+
)
|
102 |
+
|
103 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
|
104 |
+
|
105 |
+
filepaths_txt = dl_manager.download_and_extract('plant_genome_file_names.txt')
|
106 |
+
with open(filepaths_txt) as f:
|
107 |
+
filepaths = [line.rstrip() for filepath in f]
|
108 |
+
|
109 |
+
test_paths = filepaths[-2:] # 2 genomes for test set
|
110 |
+
validation_paths = filepaths[-4:-2] # 2 genomes for validation set
|
111 |
+
train_paths = filepaths[:-4] # 44 genomes for training
|
112 |
+
|
113 |
+
train_downloaded_files = dl_manager.download_and_extract(train_paths)
|
114 |
+
test_downloaded_files = dl_manager.download_and_extract(test_paths)
|
115 |
+
validation_downloaded_files = dl_manager.download_and_extract(validation_paths)
|
116 |
+
|
117 |
+
return [
|
118 |
+
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"files": train_downloaded_files, "chunk_length": self.config.chunk_length}),
|
119 |
+
datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"files": validation_downloaded_files, "chunk_length": self.config.chunk_length}),
|
120 |
+
datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"files": test_downloaded_files, "chunk_length": self.config.chunk_length}),
|
121 |
+
]
|
122 |
+
|
123 |
+
# method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
|
124 |
+
def _generate_examples(self, files, chunk_length):
|
125 |
+
key = 0
|
126 |
+
for file in files:
|
127 |
+
with open(file, 'rt') as f:
|
128 |
+
fasta_sequences = SeqIO.parse(f, 'fasta')
|
129 |
+
|
130 |
+
for record in fasta_sequences:
|
131 |
+
|
132 |
+
# parse descriptions in the fasta file
|
133 |
+
sequence, description = str(record.seq), record.description
|
134 |
+
|
135 |
+
# clean chromosome sequence
|
136 |
+
sequence = clean_sequence(sequence)
|
137 |
+
seq_length = len(sequence)
|
138 |
+
|
139 |
+
# split into chunks
|
140 |
+
num_chunks = (seq_length - 2 * self.config.overlap) // chunk_length
|
141 |
+
|
142 |
+
if num_chunks < 1:
|
143 |
+
continue
|
144 |
+
|
145 |
+
sequence = sequence[:(chunk_length * num_chunks + 2 * self.config.overlap)]
|
146 |
+
seq_length = len(sequence)
|
147 |
+
|
148 |
+
for i in range(num_chunks):
|
149 |
+
# get chunk
|
150 |
+
start_pos = i * chunk_length
|
151 |
+
end_pos = min(seq_length, (i+1) * chunk_length + 2 * self.config.overlap)
|
152 |
+
chunk_sequence = sequence[start_pos:end_pos]
|
153 |
+
|
154 |
+
# yield chunk
|
155 |
+
yield key, {
|
156 |
+
'sequence': chunk_sequence,
|
157 |
+
'description': description,
|
158 |
+
'start_pos': start_pos,
|
159 |
+
'end_pos': end_pos,
|
160 |
+
}
|
161 |
+
key += 1
|