EpicNet / README.md
gokulalex's picture
Update README.md
4ea1bc2 verified
# Dataset Summary
- The EpicNet dataset is a curated integration of peptide and protein epitope records extracted from the Immune Epitope Database (IEDB) and related immunoinformatics resources. It standardizes linear peptide epitope representations, cross-links source molecules, organisms, and taxonomy references, and unifies metadata under consistent ontological mappings.
- It supports bioinformatics, machine learning on epitope prediction, and immunotherapy research by providing aligned, structured representations of epitopes, their parent proteins, and associated immune contexts.
# Supported Tasks and Use Cases
- Epitope–Protein Mapping — link and analyze how antigenic peptides relate to their source proteins and organisms.
- Sequence-Based Epitope Prediction — train models to predict immunogenic peptide segments.
- Antigen Similarity & Cross-reactivity Studies — compare epitopes across species for vaccine development.
- Knowledge Graph Integration — generate RDF/OWL graphs for ontology-powered immunoinformatics.
- Embedding Generation and Representation Learning using transformers or peptide encoders (e.g., ESM, ProtBERT).
# Dataset Structure
## Record Count
- 2,308,224 rows representing peptide–protein associations
- Average peptide length: 8–35 amino acids
- Approximate dataset size: 85.8 MB (raw TSV) / 156 MB (Parquet)
# Data Sources and Provenance
- IEDB (Immune Epitope Database) — canonical epitope and antigen relationships
- UniProt / NCBI Protein — sequence-level references
- NCBI Taxonomy — standardized organism identifiers
- Each entry maps identifiers to IEDB IRIs, providing internationally traceable cross‑links.
# Languages and Data Format
- Sequence representation: Amino acid single-letter codes (A–Y)
- Metadata fields: English-language labels and ontology-class terms
- Format: TSV / Parquet (CSV-compatible with structured metadata)
# Intended Uses
- Computational immunology pipelines (e.g., T-cell / B-cell epitope classifier fine-tuning)
- Graph-based representation of immune networks
- Comparative peptide analytics across pathogens and hosts
- Foundation datasets for large-scale protein–epitope pretraining
# Limitations and Ethical Considerations
- Dataset reflects experimentally or computationally curated peptide sequences; not all records are experimentally verified.
- Certain peptide segments appear across species due to homologs or predicted analogs.
- Users must verify NCBI Taxonomy and UniProt IDs for clinical or diagnostic applications.
- Dataset does not contain personally identifiable data or clinical metadata.
# Citation
- If you use this dataset, please cite:
- Gokul Alex (2025). EpicNet – International Epitope Database Peptide Network. Hugging Face Datasets.
- Available at: https://huggingface.co/datasets/gokulalex/EpicNet
# License
- Creative Commons Attribution 4.0 International (CC BY‑4.0)
# Dataset Creation and Maintenance
- Author: Gokul Alex
- Release Date: October 2025
- Version: v1.0
- Contact: huggingface.co/gokulalex
# Example Entry
json
{
"IEDB IRI": "http://www.iedb.org/epitope/1",
"Object Type": "Linear peptide",
"Name": "AA + MCM(A1,A2)",
"Modified Residue(s)": "A1,A2",
"Modifications": "Main chain modification",
"Starting Position": 200,
"Ending Position": 201,
"Source Molecule": "Streptokinase",
"Source Molecule IRI": "https://uniprot.org/uniprot/P10520",
"Source Organism": "Streptococcus pyogenes",
"Species": "Streptococcus pyogenes serotype M3 D58"
}
# Acknowledgments
- Data derived from the IEDB Consortium, NCBI, and UniProt KB.
- Compilation, integration, and ontology alignment under the EpicNet Data Initiative.