Add dataset loading support

README.md

@@ -0,0 +1,76 @@
+# ToxiMol Benchmark Dataset
+
+The ToxiMol Benchmark Dataset is a collection of molecules from various toxicity datasets, designed for toxicity prediction and molecular structure repair tasks.
+
+## Dataset Description
+
+This dataset contains molecules from 11 different toxicity datasets:
+
+1. **Ames** - Molecules with Ames mutagenicity test results
+2. **Carcinogens_Lagunin** - Carcinogenic molecules from the Lagunin dataset
+3. **ClinTox** - Molecules with clinical toxicity data
+4. **DILI** - Molecules associated with drug-induced liver injury
+5. **hERG** - Molecules with hERG channel inhibition data
+6. **hERG_Central** - Alternative hERG dataset
+7. **hERG_Karim** - hERG data from Karim et al.
+8. **LD50_Zhu** - Molecules with acute toxicity (LD50) data
+9. **Skin_Reaction** - Molecules associated with adverse skin reactions
+10. **Tox21** - Molecules from the Tox21 dataset (nuclear receptor and stress response pathways)
+11. **ToxCast** - Molecules from the ToxCast dataset with pathway-specific toxicity data
+
+Each dataset entry includes:
+- Task identifier
+- Molecule ID
+- SMILES string representation of the molecule
+- Molecular structure image (PNG format)
+
+## Usage
+
+You can load the dataset using the Hugging Face `datasets` library:
+
+```python
+from datasets import load_dataset
+
+# Load a specific subdataset
+ames_dataset = load_dataset("treasurels/ToxiMol-benchmark", "ames")
+
+# Load another subdataset
+tox21_dataset = load_dataset("treasurels/ToxiMol-benchmark", "tox21")
+
+# Access the data
+for example in ames_dataset['train']:
+    print(f"Task: {example['task']}")
+    print(f"ID: {example['id']}")
+    print(f"SMILES: {example['smiles']}")
+    print("-" * 50)
+```
+
+## Dataset Structure
+
+Each subdataset follows the same structure:
+```
+{
+  "task": string,  # The toxicity task identifier
+  "id": int,       # Molecule ID
+  "smiles": string # SMILES representation of the molecule
+}
+```
+
+The molecular structure images are available in the repository in their respective subdirectories.
+
+## Citation
+
+If you use this dataset in your research, please cite:
+
+```
+@misc{toxiMol2023,
+  author = {Lin, Jason},
+  title = {ToxiMol: A benchmark dataset for toxicity prediction and molecular structure repair},
+  year = {2023},
+  publisher = {GitHub},
+}
+```
+
+## License
+
+[License information] 

dataset_infos.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ade559ae2c43c3fe340d531736a1b8d9454a52bd950335f085feca73c0e40793
+size 10290

toxiMol.py

@@ -0,0 +1,106 @@
+"""ToxiMol dataset: A benchmark dataset for toxicity prediction and molecular structure repair."""
+
+import json
+import os
+
+import datasets
+
+
+_CITATION = """
+@misc{toxiMol2023,
+  author = {Lin, Jason},
+  title = {ToxiMol: A benchmark dataset for toxicity prediction and molecular structure repair},
+  year = {2023},
+  publisher = {GitHub},
+}
+"""
+
+_DESCRIPTION = """
+ToxiMol is a benchmark dataset for toxicity prediction and molecular structure repair.
+It contains molecules from various toxicity datasets, including Ames mutagenicity, 
+carcinogenicity, clinical toxicity, drug-induced liver injury, cardiotoxicity (hERG), 
+skin reactions, acute toxicity (LD50), and other pathway-specific toxicity endpoints.
+"""
+
+_HOMEPAGE = "https://huggingface.co/datasets/treasurels/ToxiMol-benchmark"
+
+# Define the names of all the subdatasets
+_SUBDATASETS = [
+    "ames",
+    "carcinogens_lagunin",
+    "clintox",
+    "dili",
+    "herg",
+    "herg_central",
+    "herg_karim",
+    "ld50_zhu",
+    "skin_reaction",
+    "tox21",
+    "toxcast",
+]
+
+# Define the structure of each subdataset
+_FEATURES = {
+    "task": datasets.Value("string"),
+    "id": datasets.Value("int32"),
+    "smiles": datasets.Value("string"),
+}
+
+
+class ToxiMolConfig(datasets.BuilderConfig):
+    """BuilderConfig for ToxiMol."""
+
+    def __init__(self, **kwargs):
+        """BuilderConfig for ToxiMol.
+        
+        Args:
+          **kwargs: keyword arguments forwarded to super.
+        """
+        super(ToxiMolConfig, self).__init__(**kwargs)
+
+
+class ToxiMol(datasets.GeneratorBasedBuilder):
+    """ToxiMol benchmark dataset for toxicity prediction and molecular structure repair."""
+
+    BUILDER_CONFIGS = [
+        ToxiMolConfig(
+            name=subdataset,
+            version=datasets.Version("1.0.0"),
+            description=f"ToxiMol {subdataset} dataset",
+        )
+        for subdataset in _SUBDATASETS
+    ]
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(_FEATURES),
+            supervised_keys=None,
+            homepage=_HOMEPAGE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        """Returns SplitGenerators."""
+        # This dataset doesn't have predefined splits, so we use a single split
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={"subdataset": self.config.name},
+            ),
+        ]
+
+    def _generate_examples(self, subdataset):
+        """Yields examples."""
+        # Path to the JSON file for the selected subdataset
+        json_path = os.path.join(subdataset, f"{subdataset}.json")
+        
+        with open(json_path, "r", encoding="utf-8") as f:
+            data = json.load(f)
+            
+        for i, item in enumerate(data):
+            yield i, {
+                "task": item["task"],
+                "id": item["id"],
+                "smiles": item["smiles"],
+            }