Token Classification
Transformers
PyTorch
Safetensors
English
bert
biology
protein structure
token classification
Eval Results (legacy)
Instructions to use PDBEurope/BiomedNLP-PubMedBERT-ProteinStructure-NER-v1.2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use PDBEurope/BiomedNLP-PubMedBERT-ProteinStructure-NER-v1.2 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("token-classification", model="PDBEurope/BiomedNLP-PubMedBERT-ProteinStructure-NER-v1.2")# Load model directly from transformers import AutoTokenizer, AutoModelForTokenClassification tokenizer = AutoTokenizer.from_pretrained("PDBEurope/BiomedNLP-PubMedBERT-ProteinStructure-NER-v1.2") model = AutoModelForTokenClassification.from_pretrained("PDBEurope/BiomedNLP-PubMedBERT-ProteinStructure-NER-v1.2", device_map="auto") - Notebooks
- Google Colab
- Kaggle
| Feature | Description |
|---|---|
| Name | BiomedNLP-PubMedBERT-ProteinStructure-NER-v1.2 |
| Default Pipeline | transformer, ner |
| Components | transformer, ner |
| Vectors | 0 keys, 0 unique vectors (0 dimensions) |
| Sources | n/a |
| License | n/a |
| Author | Melanie Vollmar |
Label Scheme
View label scheme (19 labels for 1 components)
| Component | Labels |
|---|---|
ner |
"chemical", "complex_assembly", "evidence", "experimental_method", "gene", "mutant", "oligomeric_state", "protein", "protein_state", "protein_type", "ptm", "residue_name", "residue_name_number", "residue_number", "residue_range", "site", "species", "structure_element", "taxonomy_domain" |
Scores for entity types
| entity type | precision | recall | F1 | sample number |
|---|---|---|---|---|
| "chemical" | 0.84 | 0.90 | 0.87 | 194 |
| "complex_assembly" | 0.85 | 0.76 | 0.80 | 51 |
| "evidence" | 0.74 | 0.76 | 0.75 | 106 |
| "experimental_method" | 0.77 | 0.75 | 0.76 | 116 |
| "gene" | 0.86 | 0.92 | 0.89 | 74 |
| "mutant" | 0.83 | 0.92 | 0.88 | 258 |
| "oligomeric_state" | 0.94 | 1.00 | 0.97 | 15 |
| "protein" | 0.91 | 0.93 | 0.92 | 463 |
| "protein_state" | 0.80 | 0.83 | 0.81 | 191 |
| "protein_type" | 0.85 | 0.84 | 0.84 | 166 |
| "ptm" | 0.88 | 0.76 | 0.81 | 29 |
| "residue_name" | 0.86 | 0.95 | 0.91 | 22 |
| "residue_name_number" | 0.99 | 0.99 | 0.99 | 341 |
| "residue_number" | 1.00 | 1.00 | 1.00 | 13 |
| "residue_range" | 1.00 | 0.80 | 0.89 | 10 |
| "site" | 0.83 | 0.82 | 0.82 | 99 |
| "species" | 0.96 | 0.98 | 0.97 | 44 |
| "structure_element" | 0.88 | 0.86 | 0.87 | 319 |
| "taxonomy_domain" | 0.95 | 0.97 | 0.96 | 79 |
Data and annotations
The dataset can be found here: https://huggingface.co/datasets/PDBEurope/protein_structure_NER_model_v1.2
Citation
Vollmar, M., Tirunagari, S., Harrus, D. et al. Dataset from a human-in-the-loop approach to identify functionally important protein residues from literature. Sci Data 11, 1032 2024 https://doi.org/10.1038/s41597-024-03841-9
- Downloads last month
- 10
Evaluation results
- NER Precisionself-reported0.870
- NER Recallself-reported0.890
- NER F Scoreself-reported0.880