ActivityDiff

Classifier-guided molecular generation using diffusion models. ActivityDiff enables precise control over molecular biological activity, including targeted activation/inhibition, cooperative multi-target modulation, and off-target toxicity mitigation for de novo drug design.

Quick Start

1. Installation

git clone https://github.com/your-username/ActivityDiff.git
cd ActivityDiff

# Create conda environment
conda env create -f environment.yaml
conda activate activitydiff

2. Download Pre-trained Models

Download the pre-trained models from Releases and place them in the project root:

• epoch_064.ckpt - Pre-trained diffusion model
• P15056.pth - P15056 activity classifier
• Q02750.pth - Q02750 activity classifier

3. Run Demo

jupyter notebook main_demo.ipynb

The demo includes:

• P15056 guided generation - Generate molecules with P15056 activity
• Q02750 fragment-based design - Generate molecules from fragments with Q02750 activity
• Visualization and analysis - View generated molecules and their properties

---

This repository is built upon the lightning-hydra-template.