ProtSATT

ProtSATT: An Advanced Protein Solubility Predictor Based on Attention Mechanism

Requirements

To set up the necessary environment for ProtSATT, follow these steps:

1. Create and Activate Conda Environment

conda create -n ProtSATT python=3.10.12
conda activate ProtSATT

2. Install Python Packages

pip install keras==3.2.1
pip install Keras-Applications==1.0.2
pip install Keras-Preprocessing==1.0.1
pip install scikit-learn==1.3.0
pip install scipy==1.11.2
pip install torch==2.0.1+cu118 torchaudio==2.0.2+cu118 torchvision==0.15.2+cu118

Usage

Feature extraction

The UniRep repository, available at https://github.com/churchlab/UniRep, contains the source code, documentation, and examples for using UniRep in your projects. Due to the installation of the tape_proteins toolkit, you can conveniently utilize UniRep as follows:

tape-embed unirep input.fasta output.npz babbler-1900 --tokenizer unirep

The ESM-2 repository, available at https://github.com/facebookresearch/esm. When you need to use ESM-2 feature vectors, you can use the utils/run_esm2.py program to convert the .fasta format file to the .json format. The ProtTrans repository, available at https://github.com/agemagician/ProtTrans. When you need to use ProtT5 feature vectors, you can use the utils/run_protT5.py program to convert the .fasta format file to the .json format. Before using it, you must download all files from https://huggingface.co/Rostlab/prot_t5_xl_uniref50/tree/main.

Once you have obtained the embedding files from UniRep, ESM-2, and ProtT5, use the utils/construct_dataset.py file to build the dataset and generate a CSV file.

Train and test

• Run the train_eColi.py file to train and validate the E. coli dataset.
• Run the train_eSOL_fold.py file to optimize the hyperparameters of the eSOL dataset.
• Run the train_eSOL.py file to train the eSOL dataset.
• Run the train_Tc.py file to train and test the TR dataset.
• Run the predict.py file to predict target.

All the results will be stored in the results.txt file.