GotEnzymes2: Prediction of Enzyme Kinetics and Thermal Properties

We've got the enzymes you need

This is the official repository for the paper "GotEnzymes2: Expanding Coverage of Enzyme Kinetics and Thermal Properties". We provide a comprehensive framework for benchmarking and optimizing prediction models for enzyme properties, along with the code used to generate the GotEnzymes2 database.

Core Features

• Comprehensive Benchmarking: We retrained and evaluated multiple state-of-the-art models (UniKP, EITLEM-Kinetics, Seq2Topt, etc.) on unified datasets.
• Optimized Models: Through systematic screening, we identified the best-performing model combinations:
  • Kinetics (kcat, Km, kcat/Km): ProtT5 + MolGen + ExtraTrees
  • Thermal Properties (Topt, Tm): ProtT5 + Seq2Topt
• Large-Scale Database GotEnzymes2 contains:
  • 10,765 species
  • 7.3 million enzymes
  • 59.6 million prediction entries
  • Covering kcat, Km, kcat/Km, Topt, and Tm

Installation

# 1. Clone this repository
git clone https://github.com/LiLabTsinghua/GotEnzymes2.git
cd GotEnzymes2

# 2. Create and activate a conda environment (recommended)
conda create -n gotenzymes2 python=3.12
conda activate gotenzymes2

# 3. Install dependencies
pip install -r requirements.txt

You can also use the scripts in this repository to reproduce the model training and benchmarking results from the paper.

Data and Resources

• GotEnzymes2 Database: https://digitallifethu.com/gotenzymes
• All Data in this Study: https://github.com/LiLabTsinghua/GotEnzymes2
• Training Datasets Sources:
  • EITLEM-Kinetics
  • DeepTM
  • TOMER
  • Meltome Atlas

How to Cite

If you use our work, please cite the following paper:

Bingxue Lyu, Ke Wu, Yuanyuan Huang, Mihail Anton, Xiongwen Li, Sandra Viknander, Danish Anwer, Yunfeng Yang, Diannan Lu, Eduard Kerkhoven, Aleksej Zelezniak, Dan Gao, Yu Chen, Feiran Li. (2025). GotEnzymes2: Expanding Coverage of Enzyme Kinetics and Thermal Properties.

Contact

• Feiran Li: feiranli@sz.tsinghua.edu.cn