Enerzyme

Towards next-generation machine learning force fields for enzymatic catalysis.

Currently supported model architectures:

Model	Type	Energy and force prediction	Charge and dipole prediction	Fully modulized	Shallow ensemble	Reference paper	Reference code
PhysNet	internal	✅	✅	✅	✅	J. Chem. Theory Comput. 2019, 15, 3678–3693	Github
SpookyNet	internal	✅	✅	✅	✅	Nat. Commun. 2021, 12(1), 7273	Github
AlphaNet	internal	✅	✅	❌	❌	arXiv:2501.07155	Github
MACE	internal	✅	✅	❌	✅	NeurIPS 2022, arXiv:2206.07697	Github
NequIP	external	✅	❌	❌	❌	Nat. Commun. 2022, 13(1), 2453	Github
XPaiNN	external	✅	❌	❌	❌	J. Chem. Theory Comput. 2024, 20, 21, 9500–9511	Github

Usage

Installation

Recommended environment for internal force fields

python==3.12.8
pip==25.0
setuptools==75.8.0
h5py==3.12.1
numpy==2.2.2
addict==2.4.0
tqdm==4.67.1
joblib==1.4.2
pandas==2.2.3
pytorch==2.5.1
scikit-learn==1.6.1
ase==3.24.0
transformers==4.48.2
torch-ema==0.3
pyyaml==6.0.2
torch-scatter==2.1.2
e3nn==0.5.5

To test PhysNet, you also need

tensorflow==2.13.0

To invoke NequIP, you need

nequip==0.6.1

To invoke XPaiNN, you need

XequiNet==0.3.6
scipy==1.11.2
pyscf==2.7.0
torch_geometric==2.5.3
pytorch-warmup==0.1.1
pydantic==1.10.12

Then install the package

pip install -e .

Training

Energy (force) / Atomic Charge / Dipole moment fitting.

enerzyme train -c <configuration yaml file> -o <output directory>

Please see enerzyme/config/train.yaml for details and recommended configurations.

Enerzyme saves the preprocessed dataset, split indices, final <configuration yaml file>, and the best/last model to the <output directory>.

Active Learning Training

Please see enerzyme/config/active_learning_train.yaml for details and recommended configurations.

Evaluation

Energy (force) / Atomic Charge / Dipole moment prediction.

enerzyme predict -c <configuration yaml file> -o <output directory> -m <model directory>

Please see enerzyme/config/predict.yaml for details.

Enerzyme reads the <model directory> for the model configuration, load the models, predict the results from all active models, save the predicted values as a pickle in the corresponding model subfolders, and report the results as a csv file in the <output directory>.

Simulation

Supported simulation types:

• Constrained optimization. See enerzyme/config/opt.yaml
• Constrained flexible scan on the distance between two atoms. See enerzyme/config/scan.yaml
• Constrained Langevin MD. See enerzyme/config/nvt_md.yaml

enerzyme simulate -c <configuration yaml file> -o <output directory> -m <model directory>

Enerzyme reads the <model directory> for the model configuration, load the models, do simulation, and report the results in the <output directory>.

Fragment Extraction

Extract fragments based on local uncertainty from the prediction

enerzyme extract -c <configuration yaml file> -o <output directory> -m <model directory>

Please see enerzyme/config/extract.yaml for details.

Data annotation

Label molecules with energies, forces and dipoles from QM calculation. Please get your QM engine's environment prepared first.

enerzyme annotate -c <configuration yaml file> -o <output directory> -t <temporary directory> -s <start index> -e <end index>

Please see enerzyme/config/annotate.yaml for details.

Bond order assignment

Guess bond orders from pdb file. Compatible with QuantumPDB, where the pdb file should be the output of cluster building and the template sdf file should be ligands.sdf.

enerzyme bond -p <pdb file> -m <output mol file> -i <output image file> -t <template sdf file>