nucleotide_transformer_downstream_tasks_multilabel.py

# 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 dataset containing the "promoter_all" and "enhancers" downstream tasks from the Nucleotide
Transformer paper."""

from typing import List

import datasets
from Bio import SeqIO

# Find for instance the citation on arxiv or on the dataset repo/website
_CITATION = """\
@article{dalla2023nucleotide,
  title={The Nucleotide Transformer: Building and Evaluating Robust Foundation Models for Human Genomics},
  author={Dalla-Torre, Hugo and Gonzalez, Liam and Mendoza-Revilla, Javier and Carranza, Nicolas Lopez and Grzywaczewski, Adam Henryk and Oteri, Francesco and Dallago, Christian and Trop, Evan and Sirelkhatim, Hassan and Richard, Guillaume and others},
  journal={bioRxiv},
  pages={2023--01},
  year={2023},
  publisher={Cold Spring Harbor Laboratory}
}
"""

# You can copy an official description
_DESCRIPTION = """\
Multilabel datasets used in the Nucleotide Transformer paper.
"""

_HOMEPAGE = "https://github.com/instadeepai/nucleotide-transformer"

_LICENSE = "https://github.com/instadeepai/nucleotide-transformer/LICENSE.md"

# The toy_classification and toy_regression are two manually created configurations
# with 5 samples in both the train and test fasta files. It is notably used in order to
# test the scripts.
_TASKS_NUM_LABELS_DTYPE = [
    ("deepstarr", 6, "float32"),
    ("toy_classification", 2, "int32"),
    ("toy_regression", 2, "float32"),
]

_SPLIT_SIZES = {
    "deepstarr": {"train": 402034, "test": 41184},
    "toy_classification": {"train": 35, "test": 35},
    "toy_regression": {"train": 25, "test": 15},
}


class NucleotideTransformerDownstreamTasksConfig(datasets.BuilderConfig):
    """BuilderConfig for The Nucleotide Transformer downstream taks dataset."""

    def __init__(
        self, *args, task: str, num_labels=int, dtype: str = "int32", **kwargs
    ):
        """BuilderConfig downstream tasks dataset.
        Args:
            task (:obj:`str`): Task name.
            **kwargs: keyword arguments forwarded to super.
        """
        super().__init__(
            *args,
            name=f"{task}",
            **kwargs,
        )
        self.task = task
        self.num_labels = num_labels
        self.dtype = dtype
        self.split_sizes = _SPLIT_SIZES[task]


class NucleotideTransformerDownstreamTasks(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.1.0")
    BUILDER_CONFIG_CLASS = NucleotideTransformerDownstreamTasksConfig
    BUILDER_CONFIGS = [
        NucleotideTransformerDownstreamTasksConfig(
            task=task, num_labels=num_labels, dtype=dtype
        )
        for (task, num_labels, dtype) in _TASKS_NUM_LABELS_DTYPE
    ]
    DEFAULT_CONFIG_NAME = "deepstarr"

    def _info(self):
        features_dict = {
            "sequence": datasets.Value("string"),
            "name": datasets.Value("string"),
        }
        labels_dict = {
            f"label_{i}": datasets.Value(self.config.dtype)
            for i in range(self.config.num_labels)
        }
        features_dict.update(labels_dict)
        features = datasets.Features(features_dict)

        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,
            # Number of sequences
            dataset_size=self.config.split_sizes,
        )

    def _split_generators(
        self, dl_manager: datasets.DownloadManager
    ) -> List[datasets.SplitGenerator]:
        train_file = dl_manager.download_and_extract(self.config.task + "/train.fna")
        test_file = dl_manager.download_and_extract(self.config.task + "/test.fna")

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN, gen_kwargs={"file": train_file}
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST, gen_kwargs={"file": test_file}
            ),
        ]

    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
    def _generate_examples(self, file):
        key = 0
        with open(file, "rt") as f:
            fasta_sequences = SeqIO.parse(f, "fasta")

            for record in fasta_sequences:

                # parse descriptions in the fasta file
                sequence, name = str(record.seq), str(record.name)
                labels = [float(label) for label in name.split("|")[1:]]

                sequence_name_dict = {
                    "sequence": sequence,
                    "name": name,
                }

                labels_dict = {
                    f"label_{i}": labels[i] for i in range(self.config.num_labels)
                }
                sequence_name_dict.update(labels_dict)
                # yield example
                yield key, sequence_name_dict
                key += 1