SPIRED

An end-to-end framework for the prediction of protein structure and fitness from single sequence

Download

1. Download SPIRED codes from github

git clone https://github.com/YinghuiChen-sky/SPIRED.git
cd SPIRED

2. Download the SPIRED / SPIRED-Fitness / SPIRED-Stab model parameters and place them in the ./model folder.

mv SPIRED.pth SPIRED-Fitness.pth SPIRED-Stab.pth ./model

3. Download ESM-3B, ESM2-650M and ESMIF checkpoints, and move checkpoints to ./model dir.

wget -c https://dl.fbaipublicfiles.com/fair-esm/models/esm2_t36_3B_UR50D.pt
wget -c https://dl.fbaipublicfiles.com/fair-esm/regression/esm2_t36_3B_UR50D-contact-regression.pt
wget -c https://dl.fbaipublicfiles.com/fair-esm/models/esm2_t33_650M_UR50D.pt
wget -c https://dl.fbaipublicfiles.com/fair-esm/regression/esm1b_t33_650M_UR50S-contact-regression.pt
wget -c https://dl.fbaipublicfiles.com/fair-esm/models/esm1v_t33_650M_UR90S_1.pt
wget -c https://dl.fbaipublicfiles.com/fair-esm/models/esm1v_t33_650M_UR90S_2.pt
wget -c https://dl.fbaipublicfiles.com/fair-esm/models/esm1v_t33_650M_UR90S_3.pt
wget -c https://dl.fbaipublicfiles.com/fair-esm/models/esm1v_t33_650M_UR90S_4.pt
wget -c https://dl.fbaipublicfiles.com/fair-esm/models/esm1v_t33_650M_UR90S_5.pt
mv esm*.pt ./model

Install Conda Environment for SPIRED on Linux

Install through environment/spired_environment.yml

conda env create -f environment/spired_environment.yml

Or you can install step by step following commands below

conda create -n spired python=3.11
conda activate spired
pip install torch==2.2.0 --index-url https://download.pytorch.org/whl/cu118
pip install torch-spline-conv torch_scatter torch_sparse torch_cluster -f https://data.pyg.org/whl/torch-2.2.0+cu118.html
pip install torch_geometric==2.2.0
pip install fair-esm 
conda install numpy pandas h5py einops biopython click numba scipy scikit-learn tqdm  tensorboard  -c conda-forge
conda config --add channels https://levinthal:paradox@conda.graylab.jhu.edu
conda install pyrosetta=2023.26

Usage

• The shell scripts support the following arguments: -i input FASTA file, -o output folder, -m model directory (default: ./model), -d device (default: cpu, e.g. cuda:0).

# run SPIRED for protein structure prediction
bash run_spired.sh -i example_spired/test.fasta -m ./model -o example_spired -d cuda:0

# run SPIRED-Fitness for mutational effect (fitness) prediction + structure prediction
bash run_spired_fitness.sh -i example_fitness/test.fasta -m ./model -o example_fitness -d cuda:0

# run SPIRED-Stab for stability change prediction + structure prediction
bash run_spired_stab.sh -i example_stab/test.fasta -m ./model -o example_stab -d cuda:0

Reference

(#: co-first author; *: corresponding author)

• Yinghui Chen#, Yunxin Xu#, Di Liu, Yaoguang Xing and Haipeng Gong*. An end-to-end framework for the prediction of protein structure and fitness from single sequence. Nature Communications, 15(1):7400, 2024. DOI:10.1038/s41467-024-51776-x
• Tianyu Mi, Nan Xiao and Haipeng Gong*. GDFold2: A fast and parallelizable protein folding environment with freely defined objective functions. Protein Science, 34(2):e70041, 2025. DOI:10.1002/pro.70041

Link

• SPIRED-Fitness Server
• SPIRED-Stab Server
• Original SPIRED-Fitness Github
• If you are interested in training codes and training data examples of SPIRED/SPIRED-Fitness/SPIRED-Stab, you can follow train_code dir at Original SPIRED-Fitness Github