| tags: | |
| - protein | |
| - biology | |
| - kinase | |
| pipeline_tag: text-classification | |
| # Kinase–Substrate Interaction Classifier | |
| Fine-tuned from TAPE ProteinBERT (`bert-base`) on kinase–substrate pairs. | |
| ## Input | |
| Concatenated kinase + `<sep>` + substrate amino acid sequences. | |
| ## Output | |
| Binary classification: 1 = interaction, 0 = no interaction | |
| ## SETUP ENV | |
| ``` | |
| pip install tensorboard | |
| pip install einops | |
| pip install scikit-learn | |
| pip install plotnine | |
| pip install seaborn | |
| pip install tape-proteins pandas numpy pyyaml tqdm scikit-learn tensorboard | |
| pip install transformers | |
| pip install torch | |
| ``` | |
| ## Inference | |
| ``` | |
| from tape import ProteinBertForSequenceClassification, TAPETokenizer | |
| model = ProteinBertForSequenceClassification.from_pretrained("steveyu323/kinbert_v2_long") | |
| ``` |