PATHpre is a a Large-scale Biophysical Sampling Augmented Deep Learning model that used to predict the transition pathway of proteins.
Please kindly cite: "Exploring Protein Conformational Changes Using a Large-scale Biophysical Sampling Augmented Deep Learning Strategy", Advanced Science, 2024
numpy == 1.23.1
torch == 1.12.1
python == 3.8.0
mdtraj == 1.9.9
Example: G-protein To predict one high energy state (~transition state) from two static states (pdb1 & pdb2):
python3 PathPre.py --pdb1 2lhc.pdb --pdb2 2lhd.pdb --cut 1.0The output will be a high-energy state contact map, named HES.npy, generated with a cutoff of 1.0 nm. A 1.0 nm cutoff represents Dij = 0.5 in our paper. The recommendated cutoff is 0.8 ~ 1.2 nm depending on proteins.
Example: RfaH To predict the allosteric path from two static states (pdb1 & pdb2):
python3 PathPre.py --pdb1 2oug.pdb --pdb2 2lcl.pdb --pre_path --cut 1.0The output will include three contact maps along the allosteric path, named state2.npy, state3.npy (transition state), and state4.npy, all generated with a cutoff of 1.0 nm.
Please refer to the Dataset folder for detailed information.
Please contact Qian Wang (wqq@ustc.edu.cn) for technical support.