ProtRAP-LM: Protein Relative Accessibility Prediction through protein Language Model embeddings

Introduction

We present a novel transformer-based model, ProtRAP-LM, utilizing language model embeddings as input features, to quickly and accurately predict membrane contact probability and relative accessibility for each residue of a given protein sequence.

This package provides an implementation of the membrane contact probability (MCP), relative accessible surface area (RASA), relative lipid accessibility (RLA), relative solvent accessibility (RSA), and relative buried surface area (RBSA) prediction.

Usage

Requirements

1. PyTorch
2. Python package: numpy

conda create -n ProtRAP-LM python==3.8

conda activate ProtRAP-LM

conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia

pip install numpy==1.24.2

Command

Prediction

python ProtRAP-LM.py --input_file ./example/test.fasta --output_dir ./output/ --device cpu

$input_file is the path of the input file.

$output_dir is the directory of the output file.

$device is the device used in the calculation, 'cpu' for CPU only and 'cuda:0' for GPU support.

The model weights will be downloaded into the 'models' folder automatically during the first run. Or you can manually download the weights from Release.

Server

Please try to use our server of ProtRAP-LM predictor at:

http://www.songlab.cn

It is also avaliable at Colab.

References

Wang, L.; Zhang, J.; Wang, D.; Song, C.* Membrane Contact Probability: An Essential and Predictive Character for the Structural and Functional Studies of Membrane Proteins. PLoS Comput. Biol. 2022, 18, e1009972.

Kang, K.; Wang, L.; Song, C*. ProtRAP: Predicting Lipid Accessibility Together with Solvent Accessibility of Proteins in One Run. J. Chem. Inf. Model. 2023, 63, 1058-1065.

The implementation is based on the projects:

[1] https://github.com/facebookresearch/esm#getting-started-with-this-repo-

[2] https://github.com/ComputBiophys/MCP_Predictor

[3] https://github.com/ComputBiophys/ProtRAP