MPEK: a multi-task learning based on pre-trained language model for predicting enzymatic reaction kinetic parameters

Background

MPEK is a universal tool for predicting enzyme turnover number (k_cat) and Michaelis constant (K_m) for all types of enzymes. It uses enzyme's amino acid sequence and substrate's SMILES as inputs, with pH, temperature and species information as optional inputs, and predicts k_cat and K_m based on the CGC multi-task learning framework simultaneously. MPEK will help to improve the evaluation of enzymatic efficiency and provide important theoretical support for the screening of enzymes and the mechanistic analysis of enzymatic reactions, thereby accelerating the discovery and optimization of novel target enzymes. MPEK is therefore a promising and powerful tool for advancing biocatalysis, drug discovery, metabolic engineering, and other enzyme-dependent catalytic processes.

Installation guide

Prerequisites

• OS support: Linux
• Python version: >=3.8

Dependencies

We used the following Python packages for core development. We tested on Python 3.8.

name	version
numpy	1.24.3
pandas	1.4.3
networkx	3.1
tqdm	4.65.0
torch	1.12.0
torch-cluster	1.6.0
torch-scatter	2.1.0
torch-geometric	2.2.0
torch-sparse	0.6.14
torch-spline-conv	1.2.1
transformers	4.32.1
sentencepiece	0.1.97
rdkit	2023.9.2
tensorboard	2.15.1
scikit-learn	1.2.2

Installation

Clone codes and download necessary data files

git clone https://github.com/kotori-y/mpek.git
cd mpek/MTLKcatKM
wget https://drive.google.com/file/d/1e7IOSnQfHxII4KpOUitxlyrbBy-09DPm/view?usp=sharing # or other method
tar -zxvf checkpoints.tar.gz

Dataset

The whole dataset used for MPEK can be downloaded from this link.

Pre-training and evaluating

Pre-training

#device
DEVICE='cuda' # or cpu

# datasets
TRAIN_PATH='path/to/your/train/data'
TEST_PATH='path/to/your/test/data'
VALID_PATH='path/to/your/valid/data'

# required columns name
SEQUENCE_COLUMN='Sequence' # your column name of sequence filed in table
SMILES_COLUMN='SMILES' # your column name of smiles filed in table

# auxliary columns
ORGANISM_FLAG=1 # 1 for use organsim encoing, 0 for not
ORGANISM_COLUMN='Organism' # your column name of organism filed in table

PH_FLAG=1
PH_COLUMN='pH' # your column name of pH filed in table

TEMPERATURE_FLAG=1
TEMPERATURE_COLUMN='Temperature' # your column name of temperature filed in table

# out
CHECKPOINT_DIR='path/to/save/pretrained/model'
OUT_DIR='path/to/save/your/data/of/pretraining/progress'

sh pretrain.sh $DEVICE $TRAIN_PATH $TEST_PATH $VALID_PATH $SEQUENCE_COLUMN $SMILES_COLUMN $ORGANISM_FLAG $ORGANISM_COLUMN $PH_FLAG $PH_COLUMN $TEMPERATURE_FLAG $TEMPERATURE_COLUMN $CHECKPOINT_DIR $OUT_DIR

Evaluating

#device
DEVICE='cuda' # or cpu

# data
TEAT_PATH='path/to/your/test/data'

# required columns name
SEQUENCE_COLUMN='Sequence' # your column name of sequence filed in table
SMILES_COLUMN='SMILES' # your column name of smiles filed in table

# auxliary columns
ORGANISM_COLUMN='Organism' # your column name of organism filed in table

PH_COLUMN='pH' # your column name of pH filed in table

TEMPERATURE_COLUMN='Temperature' # your column name of temperature filed in table

sh evaluate.sh $DEVICE $TEST_PATH $SEQUENCE_COLUMN $SMILES_COLUMN $ORGANISM_COLUMN $PH_COLUMN $TEMPERATURE_COLUMN

Web Server

We also provide an easy-to-use web server: http://mathtc.nscc-tj.cn/mpek.

Cite Us

@article{10.1093/bib/bbae387,
    author = {Wang, Jingjing and Yang, Zhijiang and Chen, Chang and Yao, Ge and Wan, Xiukun and Bao, Shaoheng and Ding, Junjie and Wang, Liangliang and Jiang, Hui},
    title = "{MPEK: a multitask deep learning framework based on pretrained language models for enzymatic reaction kinetic parameters prediction}",
    journal = {Briefings in Bioinformatics},
    volume = {25},
    number = {5},
    pages = {bbae387},
    year = {2024},
    month = {08},
    issn = {1477-4054},
    doi = {10.1093/bib/bbae387},
    url = {https://doi.org/10.1093/bib/bbae387},
    eprint = {https://academic.oup.com/bib/article-pdf/25/5/bbae387/58796749/bbae387.pdf},
}

Contact

If you have questions or suggestions, please contact: [email protected].
Please see the file LICENSE for details about the "MIT" license which covers this software and its associated data and documents.