import datasets from Bio import SeqIO from Bio.SeqUtils import gc_fraction from typing import Any, Dict, List, Tuple import os import gzip import re class GenomeDatasetConfig(datasets.BuilderConfig): def __init__(self,*args, num_urls: int, **kwargs): super(GenomeDatasetConfig, self).__init__(**kwargs) self.num_urls = num_urls class GenomeDataset(datasets.GeneratorBasedBuilder): VERSION = datasets.Version("1.1.0") BUILDER_CONFIG_CLASS = GenomeDatasetConfig def _info(self): return datasets.DatasetInfo( features=datasets.Features({ "DNA_id": datasets.Value("string"), "organism": datasets.Value("string"), "year": datasets.Value("string"), "region_type": datasets.Value("string"), "specific_class": datasets.Value("string"), "product": datasets.Value("string"), "sequence": datasets.Value("string"), "gc_content": datasets.Value("float"), "translation_code": datasets.Value("string"), "start_postion": datasets.Value("int32"), "end_position": datasets.Value("int32"), }) ) # def load_genome_dataset(num_urls=None): # builder_kwargs = {} # if num_urls is not None: # builder_kwargs['config'] = GenomeDatasetConfig(num_urls=num_urls) # return load_dataset("GenomeDataset", builder_kwargs=builder_kwargs) def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]: urls_filepath = dl_manager.download_and_extract('urlfile.txt') with open(urls_filepath) as urls_file: downloaded_files = [line.rstrip() for line in urls_file] num_urls = self.config.num_urls or len(downloaded_files) downloaded_files = downloaded_files[:num_urls] train_files = downloaded_files[:int(len(downloaded_files) * 0.8)] # first 80% for training test_files = downloaded_files[int(len(downloaded_files) * 0.8):] # last 20% for testing return [ datasets.SplitGenerator( name=datasets.Split.TRAIN, gen_kwargs={"filepaths": train_files} ), datasets.SplitGenerator( name=datasets.Split.TEST, gen_kwargs={"filepaths": test_files} ) ] def _generate_examples(self, filepaths: List[str]) -> Tuple[str, Dict[str, Any]]: split_regex = re.compile('-') id_ = 0 for filepath in filepaths: if filepath.endswith(".seq.gz"): with gzip.open(filepath, 'rt') as handle: for record in SeqIO.parse(handle, "genbank"): if 'molecule_type' in record.annotations and record.annotations['molecule_type'] == 'DNA': organism = record.annotations.get('organism', 'unknown') collection_date = record.annotations.get('date', 'unknown') year = split_regex.split(collection_date)[-1] if '-' in collection_date else collection_date for feature in record.features: seq = feature.extract(record.seq) gc_content = gc_fraction(seq) start_position = int(feature.location.start) end_position = int(feature.location.end) if feature.type in ['rRNA', 'tRNA','CDS','tmRNA','mRNA','mat_peptide','sig_peptide','propeptide']: region_type = 'coding' product = feature.qualifiers.get('product', ['Unknown'])[0] if feature.type == 'CDS': translation = feature.qualifiers.get('translation', ['NA'])[0] specific_class = 'Protein' else: translation = 'NA' specific_class = feature.type elif feature.type == 'regulatory': region_type = feature.type specific_class = feature.qualifiers.get('regulatory_class', ['regulatory'])[0] translation = 'NA' product = feature.qualifiers.get('product', ['NA'])[0] elif feature.type == 'gene': continue else: if 'product' in feature.qualifiers: product = feature.qualifiers.get('product')[0] region_type = 'coding' specific_class = feature.type else: product = 'NA' region_type = feature.type specific_class = 'NA' translation = 'NA' yield str(id_), { 'DNA_id': record.id, 'organism': organism, 'year': year, 'region_type': region_type, 'specific_class': specific_class, 'product': product, 'sequence': str(seq), 'gc_content': gc_content, 'translation_code': translation, 'start_postion': start_position, 'end_position': end_position } id_+= 1