# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script # contributor. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """Script for the multi-species genomes dataset. This dataset contains the genomes from 850 different species.""" from typing import List import datasets import pandas as pd from Bio import SeqIO # Find for instance the citation on arxiv or on the dataset repo/website _CITATION = """\ @article{o2016reference, title={Reference sequence (RefSeq) database at NCBI: current status, taxonomic expansion, and functional annotation}, author={O'Leary, Nuala A and Wright, Mathew W and Brister, J Rodney and Ciufo, Stacy and Haddad, Diana and McVeigh, Rich and Rajput, Bhanu and Robbertse, Barbara and Smith-White, Brian and Ako-Adjei, Danso and others}, journal={Nucleic acids research}, volume={44}, number={D1}, pages={D733--D745}, year={2016}, publisher={Oxford University Press} } """ # You can copy an official description _DESCRIPTION = """\ Dataset made of diverse genomes available on NCBI and coming from ~850 different species. Test and validation are made of 50 species each. The rest of the genomes are used for training. Default configuration "6kbp" yields chunks of 6.2kbp (100bp overlap on each side). Similarly, the "12kbp"configuration yields chunks of 12.2kbp. The chunks of DNA are cleaned and processed so that they can only contain the letters A, T, C, G and N. """ _HOMEPAGE = "https://www.ncbi.nlm.nih.gov/" _LICENSE = "https://www.ncbi.nlm.nih.gov/home/about/policies/" _CHUNK_LENGTHS = [6000, 12000] _OVERLAP = 100 def filter_fn(char: str) -> str: """ Transforms any letter different from a base nucleotide into an 'N'. """ if char in {'A', 'T', 'C', 'G'}: return char else: return 'N' def clean_sequence(seq: str) -> str: """ Process a chunk of DNA to have all letters in upper and restricted to A, T, C, G and N. """ seq = seq.upper() seq = map(filter_fn, seq) seq = ''.join(list(seq)) return seq class MultiSpeciesGenomesConfig(datasets.BuilderConfig): """BuilderConfig for The Human Reference Genome.""" def __init__(self, *args, chunk_length: int, **kwargs): """BuilderConfig for the multi species genomes. Args: chunk_length (:obj:`int`): Chunk length. **kwargs: keyword arguments forwarded to super. """ num_kbp = int(chunk_length/1000) super().__init__( *args, name=f'{num_kbp}kbp', **kwargs, ) self.chunk_length = chunk_length class MultiSpeciesGenomes(datasets.GeneratorBasedBuilder): """Genomes from 850 species, filtered and split into chunks of consecutive nucleotides. 50 genomes are taken for test, 50 for validation and 800 for training.""" VERSION = datasets.Version("1.1.0") BUILDER_CONFIG_CLASS = MultiSpeciesGenomesConfig BUILDER_CONFIGS = [MultiSpeciesGenomesConfig(chunk_length=chunk_length) for chunk_length in _CHUNK_LENGTHS] DEFAULT_CONFIG_NAME = "6kbp" def _info(self): features = datasets.Features( { "sequence": datasets.Value("string"), "description": datasets.Value("string"), "start_pos": datasets.Value("int32"), "end_pos": datasets.Value("int32"), "fasta_url": datasets.Value("string") } ) return datasets.DatasetInfo( # This is the description that will appear on the datasets page. description=_DESCRIPTION, # This defines the different columns of the dataset and their types features=features, # Homepage of the dataset for documentation homepage=_HOMEPAGE, # License for the dataset if available license=_LICENSE, # Citation for the dataset citation=_CITATION, ) def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]: urls_file = dl_manager.download_and_extract('urls.csv') urls_df = pd.read_csv(urls_file) urls = list(urls_df['URL']) test_urls = urls[-50:] # 50 genomes for test set validation_urls = urls[-100:-50] # 50 genomes for validation set train_urls = urls[:-100] # 800 genomes for training train_downloaded_files = dl_manager.download_and_extract(train_urls) test_downloaded_files = dl_manager.download_and_extract(test_urls) validation_downloaded_files = dl_manager.download_and_extract(validation_urls) return [ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"files": train_downloaded_files, "chunk_length": self.config.chunk_length}), datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"files": validation_downloaded_files, "chunk_length": self.config.chunk_length}), datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"files": test_downloaded_files, "chunk_length": self.config.chunk_length}), ] # method parameters are unpacked from `gen_kwargs` as given in `_split_generators` def _generate_examples(self, files, chunk_length): key = 0 for file in files: with open(file, 'rt') as f: fasta_sequences = SeqIO.parse(f, 'fasta') for record in fasta_sequences: # parse descriptions in the fasta file sequence, description = str(record.seq), record.description # clean chromosome sequence sequence = clean_sequence(sequence) seq_length = len(sequence) # split into chunks num_chunks = (seq_length - 2 * _OVERLAP) // chunk_length if num_chunks < 1: continue sequence = sequence[:(chunk_length * num_chunks + 2 * _OVERLAP)] seq_length = len(sequence) for i in range(num_chunks): # get chunk start_pos = i * chunk_length end_pos = min(seq_length, (i+1) * chunk_length + 2 * _OVERLAP) chunk_sequence = sequence[start_pos:end_pos] # yield chunk yield key, { 'sequence': chunk_sequence, 'description': description, 'start_pos': start_pos, 'end_pos': end_pos, 'fasta_url': file.split('::')[-1] } key += 1