🧪 ChemEnzy-RetroPlanner Metadata Repository

This repository hosts the model metadata and environment packages for the ChemEnzy-RetroPlanner platform — a hybrid organic–enzymatic synthesis planning system integrating multi-step retrosynthesis, enzyme recommendation, and active-site annotation.

The files here are required to fully reproduce the computational environment and pretrained models for the system’s various modules.

📦 Repository Contents

File Name	Description	Size
`retro_planner_env.tar.gz`	Prebuilt conda environment for the ChemEnzyRetroPlanner	3.6 GB
`parrot_env.tar.gz`	Prebuilt conda environment for the reaction condition predictor (Parrot)	5.7 GB
`building_block_dataset.zip`	Chemical building block dataset	401 MB
`condition_predictor_metadata.zip`	Metadata for the reaction condition predictor (RCR)	144 MB
`easifa_metadata.zip`	Metadata for the enzyme active site annotator (EasIFA)	2.5 GB
`enzyme_cls_metadata.zip`	Metadata for enzyme classification module (Enzymatic Reaction Identification Model & Enzyme Recommender)	25.8 MB
`graph_retrosyn_metadata.zip`	Graph-based single-step retrosynthesis metadata (Trained on USPTO-all-remapped)	1.3 GB
`onmt_metadata.zip`	Template-based retrosynthesis metadata (BioNavi-NP)	549 MB
`rxn_filter_metadata.zip`	Reaction filtering model metadata (Reaction Plausibility Evaluator)	31 MB
`value_fun_metadata.zip`	Value function for synthesis planning (Depth-based guiding function model)	2 MB
`USPTO_condition.mar`	Trained reaction condition predictor (Parrot)	101 MB

⚙️ Quick Installation Guide

1️⃣ Clone and Configure

git clone https://github.com/wangxr0526/ChemEnzyRetroPlanner.git
cd ChemEnzyRetroPlanner
chmod +x ./setup_ChemEnzyRetroPlanner.sh
./setup_ChemEnzyRetroPlanner.sh

Alternatively, you may use the provided auto-installation script (see below).

🧰 Automatic Setup Script

The repository includes a Bash script to automatically:

Configure environment variables
Download and unpack all required .tar.gz and .zip files
Verify model existence before download
Install ESM-2 pretrained weights (for enzyme feature extraction)
Prepare the conda environment for immediate use

To execute:

bash setup_environment.sh

The script will:

Create necessary directories under $CONDA_HOME/envs
Unpack the prebuilt conda environment retro_planner_env.tar.gz
Activate the environment
Download and place metadata under ./metadata/
Retrieve structures.csv.gz from ChEBI for EasIFA integration

🧾 Dependencies

Conda ≥ 23.0
gdown, curl, wget, tar, unzip

All dependencies are pre-packaged within retro_planner_env.tar.gz.

🪄 Notes

Some .zip files contain large model weights; ensure sufficient disk space (>15 GB).
Certain files (e.g., onmt_metadata.zip) may be flagged as unsafe due to pickle serialization — these are safe if used within the provided environment.

📬 Contact

For technical support or collaborative inquiries: 📧 wangxr2018@lzu.edu.cn 🧩 Maintainer: Xiaorui Wang, Postdoctoral Research Fellow in Zhejiang University

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