| AuroBind Checkpoint | AuroFast Checkpoint |

AuroBind is a fitness-aligned structural modeling method designed for high-throughput virtual screening. It takes protein sequences and small molecule SMILES as direct inputs to predict their complex's fitness score, while also supporting structure prediction and confidence scoring.

We provide two models for inference: the standard AuroBind and the lightweight AuroFast. An A100 80GB or higher-memory GPU is recommended for standard model

1. Prepare Input File: Create a YAML file with your sequences following following our input format specification

2. Download Cache Data and Model: You can download from google drive. Place the downloaded files in the src/cache_data/ directory before running inference.

3. Installation and demo:

To more complete installation instructions and usage, please refer to the Installation Guide.

bash predict.sh

4. Output:Predictions will be saved to: ./output/5S8I_A/predictions

We provide a Jupyter Notebook demo for using AuroFast to perform inference on 10 targets.

1. Download Required Files: To run the demo, Please download the demo trunk here, and AuroFast model checkpoints here.

2. File Placement: Unzip the downloaded files and place them in the src/aurofast folder.

3. Run Demo: Open and execute

inference_aurofast_demo.ipynb

This will generate predictions for the input file demo_input.csv, and the results will be saved in inference_result.csv. To obtain additional features_fast_X.pt files, please run the full AuroBind pipeline.

We highly recommend using the AuroBind Server for user-friendly and accurate protein-ligand structure and fitness predictions. It requires no installation and provides an intuitive web interface that lets you submit your sequences and visualize results directly in your browser.

• The implementation of fast layernorm operators is inspired by OneFlow and FastFold, following Protenix's usage.
• Many components are adapted from OpenFold, with substantial modifications and improvements by our team (except for the LayerNorm part).
• This repository implements the Inference Data Pipeline (including data/feature processing and MSA generation) drawing conceptual inspiration and selectively reusing code components from IntelliFold.

This code repository is licensed under the Creative Commons Attribution-Non-Commercial ShareAlike International License, Version 4.0 (CC-BY-NC-SA 4.0) (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at https://github.com/GENTEL-lab/AuroBind/blob/main/LICENSE.