molann 1.1.7

"PyTorch Artificial Neural Networks (ANNs) for Molecular Systems",

MolANN

This package implements classes that define PyTorch Artificial Neural Network (ANN) functions taking into account alignment (translation and rotation), and functions that depend on molecular features (bonds, angles, etc.).

Installation

The package can be installed via pip:

pip install molann

The installation from source is described in the Installation page.

Simple example

The following code defines an ANN as a function of a bond distance and a dihedral angle.

import MDAnalysis as mda
from molann.ann import FeatureLayer, PreprocessingANN, MolANN, create_sequential_nn
from molann.feature import Feature

# pdb file of the system
pdb_filename = '/path/to/system.pdb'
ref = mda.Universe(pdb_filename)

# define a feature that describes the bond between atoms 5 and 6.
f1 = Feature('name', 'bond', ref.select_atoms('bynum 5 6'))
# define a feature that describes the dihedral angle formed by the first 4 atoms.
f2 = Feature('name', 'dihedral', ref.select_atoms('bynum 1 3 2 4'))

# define feature layer using the above two features.
f_layer = FeatureLayer([f1,f2], use_angle_value=False)

# define the preprocessing layer.
# we do not need alignment, since both features f1 and f2 are translation- and rotation-invariant.
pp_layer = PreprocessingANN(None, f_layer)

output_dim = pp_layer.output_dimension()
# define neural network layers that contain training parameters
nn = create_sequential_nn([output_dim, 5, 3])

# define the final network
model = MolANN(pp_layer, nn)

torch.jit.script(model).save('model.pt')

More examples for each class can be found in the MolANN docs .

Documentataion

Please refer to MolANN docs.