GDFold2 is a protein folding environment. It is designed to rapidly and parallelly fold the protein structures based on arbitrary predicted constraints, which could be freely integrated into the environment as user-defined loss functions. We provide four folding modes to match the different geometric information. You can also customize the constraints according to your specific needs.
git clone https://github.com/Gonglab-THU/GDFold2.git
cd GDFold2conda env create -f environment.yml
conda activate GDFold2-
fold.py: input protein sequence (.fasta format) and predicted geometric information (.npz format) and output protein structure(s).python fold.py example/test.fasta example/test.npz example -d cuda
-
Please install PyRosetta first!
-
relax.py: perform FastRelax procedure.python relax.py --input example/101M_1.pdb --output example/relax.pdb
-
QAmodel/run.py: input a directory containing multiple protein models folded by GDFold2 and output their ranking filerank.txtin the input directory.python QAmodel/run.py --input QAmodel/example
-
Step 1: run
Dynamics/pdb2cst.pyto convert two conformational states of the same protein target into geometric constraint file (comb.npz).python Dynamics/pdb2cst.py --state1 Dynamics/1ake_A.pdb --state2 Dynamics/4ake_A.pdb --output Dynamics
-
Step 2: run
fold.pyto predict the possible conformations in the transition path between the two conformational states.python fold.py Dynamics/comb.fasta Dynamics/comb.npz Dynamics/dynamics -n 50 -m Dynamics -d cuda
We provide a web sever (GDFold2) for exploring protein structural dynamics. You can copy all the characters from Dynamics/1ake_A.pdb and Dynamics/4ake_A.pdb and paste them separately into the input box of the web server for testing.
If you use this code in your research, please cite our paper:
@article
author = {Mi, Tianyu and Gong, Haipeng},
title = {GDFold2: a fast and parallelizable protein folding environment with freely defined objective functions},
year = {2024},
doi = {10.1101/2024.03.13.584741}