Description

This repository contains the code and datasets to reproduce the results and figures and to train the models from our manuscript "A general substrate prediction model for transport proteins reveals differences between transporter types and localisations".

For people interested in using the trained prediction model, we implemented a web server that allows an easy use of our trained model. The prediction tool can be run in a web-browser and does not require the installation of any software. Prediction results are usually ready within a few minutes.

Downloading data folder

Before you can run all scripts of this repository, you need to download and unzip an additional data folder from Zenodo. Afterwards, this repository should have the following strcuture:

├── code
├── data
├── S1_data.xlsx         
├── S2_data.xlsx         
└── README.md

Requirements for running the code

I suggest to install a new conda environment with all required python packages:

Creating new conda environment:

conda create -n Transporter python=3.7

activate environment

conda activate Transporter

Install jupyter notebook

conda install jupyter jupyter notebook

Installing python packages

conda create -n Transport python=3.7 conda activate Transport conda install jupyter conda install pandas pip install goatools pip install PubChemPy pip install biopython pip install openpyxl pip install bioservices conda install -c rdkit rdkit pip install scikit-learn

conda install -c conda-forge py-xgboost=1.3.3 pip install torch