HamNet: Conformation-Guided Molecular Representation with Hamiltonian Neural Networks

TL;DR

HamNet introduces a novel Hamiltonian neural network that effectively preserves 3D molecular conformations, leading to state-of-the-art molecular property predictions on benchmark datasets.

Contribution

The paper presents a new molecular representation method using Hamiltonian neural networks to incorporate 3D conformations into molecular fingerprints.

Findings

HamNet accurately preserves molecular conformations.

Fingerprints from HamNet outperform existing methods on MoleculeNet.

The Hamiltonian Engine effectively models atomic interactions in molecules.

Abstract

Well-designed molecular representations (fingerprints) are vital to combine medical chemistry and deep learning. Whereas incorporating 3D geometry of molecules (i.e. conformations) in their representations seems beneficial, current 3D algorithms are still in infancy. In this paper, we propose a novel molecular representation algorithm which preserves 3D conformations of molecules with a Molecular Hamiltonian Network (HamNet). In HamNet, implicit positions and momentums of atoms in a molecule interact in the Hamiltonian Engine following the discretized Hamiltonian equations. These implicit coordinations are supervised with real conformations with translation- & rotation-invariant losses, and further used as inputs to the Fingerprint Generator, a message-passing neural network. Experiments show that the Hamiltonian Engine can well preserve molecular conformations, and that the fingerprints generated by HamNet achieve state-of-the-art performances on MoleculeNet, a standard molecular machine learning benchmark.