Hugging Face's logo

jglaser
/
protein-ligand-mlp-1

Sentence Similarity
sentence-transformers
feature-extraction
Inference Endpoints
Model card Files Community
protein-ligand-mlp-1 / README.md
jglaser's picture
jglaser
Update README.md
c04b888
preview code
|
raw
history blame contribute delete
3.38 kB
metadata
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - feature-extraction
  - sentence-similarity

This model is superseded by https://github.com/ORNL/affinity_pred

jglaser/protein-ligand-mlp-1

This is a sentence-transformers model: It maps pairs of protein and chemical sequences (canonical SMILES) onto binding affinities (pIC50 values).

Each member of the ensemble has been trained using a different seed and you can use the different models as independent samples to estimate the uncertainty.

Usage (Sentence-Transformers)

Using this model becomes easy when you have sentence-transformers installed:

#pip install -U sentence-transformers
pip install git+https://github.com/jglaser/sentence-transformers.git@enable_mixed

Then you can use the model like this:

from sentence_transformers import SentenceTransformer
sentences = [{'protein': ["SEQVENCE"], 'ligand': ["c1ccccc1"]}]

model = SentenceTransformer('jglaser/protein-ligand-mlp-1')
embeddings = model.encode(sentences)
print(embeddings)

Evaluation Results

Full Model Architecture 

SentenceTransformer(
  (0): Asym(
    (protein-0): Transformer({'max_seq_length': 2048, 'do_lower_case': False}) with Transformer model: BertModel 
    (protein-1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
    (protein-2): Dense({'in_features': 1024, 'out_features': 1024, 'bias': True, 'activation_function': 'torch.nn.modules.activation.Tanh'})
    (ligand-0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
    (ligand-1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
    (ligand-2): Dense({'in_features': 768, 'out_features': 768, 'bias': True, 'activation_function': 'torch.nn.modules.activation.Tanh'})
  )
  (1): Dense({'in_features': 1792, 'out_features': 1000, 'bias': True, 'activation_function': 'torch.nn.modules.activation.GELU'})
  (2): Dense({'in_features': 1000, 'out_features': 1000, 'bias': True, 'activation_function': 'torch.nn.modules.activation.GELU'})
  (3): Dense({'in_features': 1000, 'out_features': 1000, 'bias': True, 'activation_function': 'torch.nn.modules.activation.GELU'})
  (4): Dense({'in_features': 1000, 'out_features': 1, 'bias': True, 'activation_function': 'torch.nn.modules.linear.Identity'})
  (5): Dense({'in_features': 1, 'out_features': 1, 'bias': True, 'activation_function': 'torch.nn.modules.linear.Identity'})
)

Citing & Authors

Find more information in our bioRxiv preprint