Biophysical Representation of Interfaces via Delaunay-based Graph Embeddings (BRIDGE)

BRIDGE is a lightweight, cutting-edge coarse-grained molecular AI model designed for extracting biophysically meaningful embeddings from protein-protein interfaces by leveraging Delaunay tessellation.

Citation

 BRIDGE: A Coarse-Grained Architecture to Embed Protein-Protein Interactions for Therapeutic Applications
Pranav M Khade, Andrew M Watkins
bioRxiv 2025.04.11.648307; doi: https://doi.org/10.1101/2025.04.11.648307 

Prerequisites

All the requirements for training the BRIDGE encoder and applying it for Viscosity and/ Affinity prediction tasks through scripts can be found here.

Building the BRIDGE 🌉

BRIDGE model can be trained using the train.py file. The following are the parameters for training. Also check config.yaml for the hyperparameters.

usage: train.py [-h] [--session SESSION] [--logger {tensorboard,wandb}] [--nodes NODES] [--lr LR] [--epochs EPOCHS] [--train_batchsize TRAIN_BATCHSIZE]
                [--valid_batchsize VALID_BATCHSIZE] [--test_batchsize TEST_BATCHSIZE] [--seed SEED]

options:
  -h, --help            show this help message and exit
  --session SESSION     Session directory (default: default)
  --logger {tensorboard,wandb}
                        Logger. (default: tensorboard)
  --nodes NODES
  --lr LR               Learning rate (default: 0.003)
  --epochs EPOCHS
  --train_batchsize TRAIN_BATCHSIZE
  --valid_batchsize VALID_BATCHSIZE
  --test_batchsize TEST_BATCHSIZE
  --seed SEED

Example command to train

python3 train.py --epochs 300 --lr 0.0001

BRIDGE Encodings Application

BRIDGE Encodings can be used for many tasks related to proteins. However, we have applied it to Affinity and Viscosity of Antibodies in the manuscript. Please check the scripts/Affinity.py and scripts/Viscosity.py files' main functions that list all the utilities with appropriate names such as:

• create_vectorized_data_from_pdb(): Generate and store embedding for user-specified files.
• train() & test(): Train & test the models for specific tasks respectively
• BRIDGE_UMAP(): Dimensionality reduction analysis of entire data un BRIDGE UMAP latest space.

Users are encouraged to look into single_entry_process() to create custom functions for their own data and tasks.

Viscosity Leave One Out Cross Validation (LOOCV)

To run the viscosity analysis from the manuscript, please run the following command:

wandb sweep --project gnn_interface_viscosity_pred misc/viscosity_loo_sweep.yml

Follow the instructions given by wandb after running the command.

Data

All the data used for the affinity & viscosity model is present in the data folder. Please give proper credit to the original authors if you use their data. Following is the list of datasets and links to the papers:

• Sabdab Affinity
• Pierce lab affinity subset
• Viscosity Dataset

Results

All the images will be stored in images folders after running different scripts.