metadata
language: en
license: mit
datasets:
- ronig/protein_binding_sequences
Peptriever: A Bi-Encoder for large-scale protein-peptide binding search
For training details see our Application Note.
Training code can be found in our Github repo.
A live demo is available on our application page
Usage
import torch
from transformers import AutoModel, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("ronig/protein_biencoder")
model = AutoModel.from_pretrained("ronig/protein_biencoder", trust_remote_code=True)
model.eval()
peptide_sequence = "AAA"
protein_sequence = "MMM"
encoded_peptide = tokenizer.encode_plus(peptide_sequence, return_tensors='pt')
encoded_protein = tokenizer.encode_plus(protein_sequence, return_tensors='pt')
with torch.no_grad():
peptide_output = model.forward1(encoded_peptide)
protein_output = model.forward2(encoded_protein)
print("distance: ", torch.norm(peptide_output - protein_output, p=2))
Version
Model checkpint: peptriever_2023-06-23T16:07:24.508460