KcatNet: A Geometric Deep Learning Framework for Genome-Wide Prediction of Enzyme Catalytic Efficiency
We introduced KcatNet, a geometric deep learning model for high-throughput prediction of turnover number in metabolic enzymes across all organisms, leveraging paired enzyme sequence data and substrate Simplified Molecular Input Line Entry System (SMILES) representations.
KcatNet is implemented using Python 3.9.12 and various scientific computing packages (numpy, biopython, scikit-learn, torch). We recommend creating a dedicated virtual environment and installing the same package versions.
conda env create -f environment_gpu.yml
Alternatively, command lines can be helpful in setting up the environment (tested on linux with python 3.9), this can be done in few lines using Anaconda (https://www.anaconda.com).
conda create --name turnover python=3.9
conda activate turnover
conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 -c pytorch
KcatNet utilizes pre-trained protein language model protT5 and ESM for protein residual embedding and evolutionary information extraction.
(1) Install protT5
pip install transformers
(2) Install ESM
pip install "fair-esm"
Note: The language model protT5 (referring to https://github.com/agemagician/ProtTrans) can also be downloaded for offline use.
We also included a pretrained molecular language modoel SMILES Transformer to generate substrate representation, the link is also provided on SMILES Transformer.
The script pred_kcat.py can be utilized to predict the enzyme turnover number given an enzyme sequence and substrate SMILE pair:
python pred_kcat.py
we give several enzyme-substrate pairs as example in the examples/example.xlsx, where each pair included the enzyme sequence and one substrate SMILE:
| Pro_seq | Smile | Predicted Kcats |
|---|---|---|
| MALKLNPFLSQTQKLPS.... | O=P(O)(O)OC[C@H]1O... | 0.755690265807342 |
| MVQDRKSTVVITGASSG.... | [H+] | 0.247781748919223 |
Enzyme turnover number datasets used for model development and evaluation are provided in the ./Dataset folder. The model can be Trained from scratch via:
python train.py
For any question and comment regarding the code, please reach out to tong.pan@monash.edu