HuggingLigand Binding Affinity Predictor

Model Details

Model Description

This model is a feedforward neural network trained to predict protein-ligand binding affinity values using embeddings of proteins and ligands. Protein embeddings are generated using ProtT5, and ligand embeddings using ChemBERTa. The model predicts continuous affinity metrics such as Kd, Ki, or IC50, enabling virtual screening and drug discovery tasks without requiring explicit 3D structural data.

  • Developed by: RSE Group 11
  • Model type: Feedforward neural network
  • License: MIT
  • Language: English

Model Sources

  • Repository: HuggingLigand
  • Dataset used for training: [Hugging Face dataset link to RSE-Group11/Hugging-Ligand-embeddings]

Uses

Direct Use

Use this model to estimate protein-ligand binding affinity from precomputed embeddings for downstream drug discovery tasks such as virtual screening and prioritizing candidate molecules.

Out-of-Scope Use

This model is not designed for predicting 3D molecular conformations or generating new molecules. It also may not generalize well outside the chemical space represented in the training data.

Bias, Risks, and Limitations

The model inherits biases present in the training data, including chemical diversity and assay variability. Predictions should be interpreted with caution and validated experimentally. The model does not account for explicit structural or dynamic effects of proteins or ligands.

How to Get Started with the Model


Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Dataset used to train RSE-Group11/hugging-ligand-affinity-predictor