Utilize automated machine learning to predict gene synthesis plant specialized metabolites

1. System requirements
2. Installation guide
3. Demo
4. Instructions for use
5. License
6. Reference

The software was tested on CentOS 7.5.1804 (Core) with Miniconda3. Other packages were used as follows:

python 3.8.12
autogluon 0.3.1
numpy
scipy
pandas

Install miniconda:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh && bash Miniconda3-latest-Linux-x86_64.sh

Create a new conda environment and update

conda create --name autogluon python==3.8.12 conda update -n base conda

Activate conda environment

conda activate autogluon

Install dependencies: pip install -U setuptools wheel

pip install autogluon==0.3.1

Install ML-pre-GSPSM:

git clone https://github.com/baiwenhuim/ML-pre-GSPSM.git

The directory including in model script is changed into:

cd /ML-pre-GSPSM/model

And you can also download the pre-trained model in three species (Arabidopsis thaliana, maize, and tomato) from Zenodo.

Once downloaded, you can unzip it:

unzip azs.zip

Then, you can use the script predict_unknown_gene.py to predict enzyme genes which you want to predict:

python predict_unknown_gene.py --path /prefix/model_path/azs --input /prefix/input_file.csv --output output.csv

The following results are output into"/Downloaded_directory_path/output.csv":

● Cross validation results in the basic models and AutoGluon-Tabular
● Test results in the AutoGluon-Tabular mode.

Run time depends on the dataset size.

This software is released under the MIT License, according to LICENSE.