hugging-science/AI4FA-Goldberg

Food Allergy Microbiome Dataset (Experimental)

Dataset Summary

This is an experimental microbiome dataset designed for exploratory research in food allergy classification. The dataset contains multiple data modalities (DNA embeddings, microbiome embeddings, raw DNA sequences) collected longitudinally at several timepoints.

Warning: This dataset is experimental. Its structure is frozen for ongoing research, and it is not ready for benchmarking.

---

Dataset Structure

The dataset is organized by data type and timepoint:

.
├── dna_embeddings
│   └── month_{1,2,3,6,12,24,36}/dna_embeddings.h5
├── dna_sequences
│   └── month_{1,2,3,6,12,24,36}/*.csv
└── microbiome_embeddings
    └── month_{1,2,3,6,12,24,36}/microbiome_embeddings.h5

• DNA embeddings: .h5 files with embedding vectors derived from DNA sequences.
• Microbiome embeddings: .h5 files containing microbiome feature vectors.
• DNA sequences: raw .csv files representing sequences or processed features.

Each timepoint contains multiple samples per subject. File names (e.g., DRS061544.csv) serve as sample IDs. Subject IDs and mappings are implicit; users must manage them carefully.

---

Intended Use

• Task: Exploratory classification of food allergies.
• Users are expected to define their own train/test splits.
• Critical: Do not split samples randomly; multiple samples per subject exist. Splits should be done at the subject level to avoid data leakage.

---

Data Notes

• Longitudinal: Samples are collected at multiple months (1, 2, 3, 6, 12, 24, 36).
• Multi-modal: Embeddings and sequences are provided separately; users may combine them as needed.
• No labels are embedded per file. Labels must be handled separately or mapped from your internal records.
• This dataset is research-focused, not benchmark-ready.

---

License

This experimental dataset is currently released under Apache License 2.0

---