Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,138 @@
+---
+configs:
+- config_name: default
+  data_files: "data/swap_complexes.parquet"
+  default: true
+license: cc-by-4.0
+task_categories:
+- other
+tags:
+- protein
+- complex
+- evolution
+- swap
+- self-supervised
+- protein-protein-interaction
+- ortholog
+size_categories:
+- 100K<n<1M
+---
+
+# Protein Complex Swap Dataset
+
+A dataset of cross-species protein-protein interaction comparisons for self-supervised training of protein complex models.
+
+## Overview
+
+Each row contains **two native protein complexes from different species** that share the same conserved interaction (via orthology). From each row, a model can construct:
+
+- **Native complex 1**: (A_sp1_seq, B_sp1_seq) — correct interaction in species 1
+- **Native complex 2**: (A_sp2_seq, B_sp2_seq) — correct interaction in species 2
+- **Swap complex 1**: (A_sp1_seq, B_sp2_seq) — cross-species partner swap
+- **Swap complex 2**: (A_sp2_seq, B_sp1_seq) — symmetric swap
+
+All **4 protein sequences** are included inline in each row, along with two quantitative distance measures.
+
+## Dataset Statistics
+
+| Metric | Value |
+|--------|-------|
+| Total rows | 142,598 |
+| Unique interaction groups | 40,973 |
+| Unique proteins | 26,775 |
+| Species covered | 9 model organisms |
+
+## Distance Measures
+
+Each row contains two quantitative measures of "how different" the swap is:
+
+1. **`seq_identity_A`**: Pairwise sequence identity between protein A in species 1 and protein A in species 2 (0.0–1.0)
+2. **`seq_identity_B`**: Pairwise sequence identity between protein B in species 1 and protein B in species 2 (0.0–1.0)
+3. **`taxonomic_distance`**: NCBI taxonomy tree distance between the two species (integer, higher = more distant)
+
+## Species Coverage
+
+| Species 1 | Species 2 | Rows |
+|-----------|-----------|------|
+| Homo sapiens | Mus musculus | 34,514 |
+| Saccharomyces cerevisiae | Homo sapiens | 12,309 |
+| Mus musculus | Rattus norvegicus | 12,061 |
+| Homo sapiens | Rattus norvegicus | 11,202 |
+| Saccharomyces cerevisiae | Mus musculus | 8,692 |
+| Drosophila melanogaster | Homo sapiens | 5,553 |
+| Arabidopsis thaliana | Homo sapiens | 5,492 |
+| Drosophila melanogaster | Mus musculus | 5,117 |
+| Arabidopsis thaliana | Mus musculus | 4,911 |
+| Arabidopsis thaliana | Saccharomyces cerevisiae | 4,402 |
+| Saccharomyces cerevisiae | Rattus norvegicus | 4,101 |
+| Caenorhabditis elegans | Homo sapiens | 4,013 |
+| Caenorhabditis elegans | Mus musculus | 3,733 |
+| Saccharomyces cerevisiae | Drosophila melanogaster | 3,481 |
+| Saccharomyces cerevisiae | Caenorhabditis elegans | 2,985 |
+| Drosophila melanogaster | Rattus norvegicus | 2,691 |
+| Arabidopsis thaliana | Rattus norvegicus | 2,570 |
+| Arabidopsis thaliana | Drosophila melanogaster | 2,389 |
+| Caenorhabditis elegans | Drosophila melanogaster | 2,251 |
+| Caenorhabditis elegans | Rattus norvegicus | 2,154 |
+| Arabidopsis thaliana | Caenorhabditis elegans | 2,023 |
+| Danio rerio | Homo sapiens | 1,579 |
+| Danio rerio | Mus musculus | 1,472 |
+| Drosophila melanogaster | Danio rerio | 718 |
+| Saccharomyces cerevisiae | Danio rerio | 611 |
+| Danio rerio | Rattus norvegicus | 599 |
+| Caenorhabditis elegans | Danio rerio | 542 |
+| Arabidopsis thaliana | Danio rerio | 433 |
+
+
+## Column Schema
+
+| Column | Type | Description |
+|--------|------|-------------|
+| `interaction_group_id` | str | Groups rows from the same conserved interaction |
+| `species_1` | str | Species of complex 1 |
+| `taxid_1` | int | NCBI taxonomy ID of species 1 |
+| `protein_A_sp1_uniprot` | str | UniProt accession of protein A in species 1 |
+| `protein_A_sp1_gene` | str | Gene name of protein A in species 1 |
+| `protein_A_sp1_seq` | str | Full amino acid sequence of protein A in species 1 |
+| `protein_A_sp1_len` | int | Sequence length |
+| `protein_B_sp1_uniprot` | str | UniProt accession of protein B in species 1 |
+| `protein_B_sp1_gene` | str | Gene name of protein B in species 1 |
+| `protein_B_sp1_seq` | str | Full amino acid sequence of protein B in species 1 |
+| `protein_B_sp1_len` | int | Sequence length |
+| `string_score_sp1` | int | STRING confidence score for (A, B) in species 1 |
+| `species_2` | str | Species of complex 2 |
+| `taxid_2` | int | NCBI taxonomy ID of species 2 |
+| `protein_A_sp2_uniprot` | str | UniProt accession of protein A in species 2 |
+| `protein_A_sp2_gene` | str | Gene name |
+| `protein_A_sp2_seq` | str | Full amino acid sequence ← model input |
+| `protein_A_sp2_len` | int | Sequence length |
+| `protein_B_sp2_uniprot` | str | UniProt accession of protein B in species 2 |
+| `protein_B_sp2_gene` | str | Gene name |
+| `protein_B_sp2_seq` | str | Full amino acid sequence ← model input |
+| `protein_B_sp2_len` | int | Sequence length |
+| `string_score_sp2` | int | STRING confidence score for (A, B) in species 2 |
+| `seq_identity_A` | float | Sequence identity between A_sp1 and A_sp2 (0.0–1.0) |
+| `seq_identity_B` | float | Sequence identity between B_sp1 and B_sp2 (0.0–1.0) |
+| `taxonomic_distance` | float | NCBI taxonomy tree distance between species 1 and 2 |
+| `role_key_A` | str | Ortholog group ID for protein A family |
+| `role_key_B` | str | Ortholog group ID for protein B family |
+
+## Construction Method
+
+1. **Protein-protein interactions**: STRING v12.0 physical interactions (score ≥ 700) for 9 model organisms
+2. **Ortholog mapping**: STRING best-hit homology to identify same proteins across species
+3. **Interaction groups**: Interactions where both partners have orthologs in ≥2 species
+4. **Sequences**: UniProt REST API batch retrieval
+5. **Taxonomic distance**: NCBI taxonomy tree edge distance
+6. **Sequence identity**: Pairwise sequence alignment
+
+## Intended Use
+
+Self-supervised training for protein complex preference models. The model learns to distinguish native complexes from cross-species swaps at varying evolutionary distances.
+
+## Citation
+
+If you use this dataset, please cite:
+
+- STRING database: Szklarczyk et al., "The STRING database in 2023", Nucleic Acids Research (2023)
+- UniProt: UniProt Consortium, "UniProt: the Universal Protein Knowledgebase in 2025", Nucleic Acids Research (2025)