AModels
AlphaFold
DeepMind's protein-structure prediction system, with open AlphaFold 2 code and a public database of predicted structures.
proteinsstructure prediction
View entry AI4S MODELS
AI4S models
Scientific machine-learning models, model families, and direct implementations for fields including life science, chemistry, and materials research. The database currently includes 11 entries in this section.
OModels
OpenFold
A trainable open-source implementation of AlphaFold 2 intended for protein-structure research and model development.
proteinsopen model
View entry GModels
GNoME
A Google DeepMind research system and dataset for predicting the stability of candidate inorganic materials with graph neural networks.
materials discoverygraph neural networks
View entry MModels
MatterGen
An open-source generative model from Microsoft Research for proposing inorganic crystal structures, including property-conditioned generation.
generative modelinorganic materials
View entry DModels
DiffDock
Research code that applies a diffusion-based generative model to predict protein-small-molecule ligand binding poses.
molecular dockingdiffusion model
View entry BModels
Boltz
An open-source family of models and inference software for biomolecular complex structure and selected binding-affinity predictions.
biomolecular structurebinding affinity
View entry RModels
RFdiffusion
Open-source research code and model weights for diffusion-based generation of protein backbones, motif scaffolds, and symmetric structures.
protein designdiffusion model
View entry NModels
NequIP
An open-source framework for training E(3)-equivariant machine-learning interatomic potentials, with integrations for ASE and LAMMPS workflows.
machine-learning potentialsequivariant networks
View entry MModels
MACE
An open-source PyTorch framework for training and deploying machine-learning interatomic potentials with higher-order equivariant message passing.
machine-learning potentialsatomistic simulation
View entry CModels
CHGNet
A pretrained graph-neural-network potential and open-source tools for charge-informed atomistic prediction and fine-tuning.
neural potentialsmaterials simulation
View entry CModels
Chemprop
An open-source machine-learning package for training message-passing neural networks on molecular and reaction property data.
molecular propertiesmessage passing
View entry