MolSpectra: Pre-training 3D Molecular Representation with Multi-modal Energy Spectra
Liang Wang, Shaozhen Liu, Yu Rong, Deli Zhao, Qiang Liu, Shu Wu, Liang Wang
2025-02-27
Summary
This paper talks about a new way to teach computers about molecules called MolSpectra. It uses energy spectra, which are like unique fingerprints of molecules, to help computers understand molecules better in 3D.
What's the problem?
Current methods for teaching computers about molecules only use classical physics, which misses out on important quantum effects. This means computers aren't getting the full picture of how molecules really behave, especially when it comes to their energy levels.
What's the solution?
The researchers created MolSpectra, which uses energy spectra to teach computers about molecules. They also made a special tool called SpecFormer that can understand these energy spectra. By combining this with 3D information about molecules, they helped computers learn a more complete and accurate picture of how molecules work.
Why it matters?
This matters because better understanding of molecules can lead to breakthroughs in chemistry, drug discovery, and materials science. By teaching computers to understand molecules more like how they actually behave in nature, including quantum effects, scientists can make more accurate predictions about molecular properties and how molecules might interact. This could speed up the process of developing new medicines or materials with specific properties.
Abstract
Establishing the relationship between 3D structures and the energy states of molecular systems has proven to be a promising approach for learning 3D molecular representations. However, existing methods are limited to modeling the molecular energy states from classical mechanics. This limitation results in a significant oversight of quantum mechanical effects, such as quantized (discrete) energy level structures, which offer a more accurate estimation of molecular energy and can be experimentally measured through <PRE_TAG>energy spectra</POST_TAG>. In this paper, we propose to utilize the <PRE_TAG>energy spectra</POST_TAG> to enhance the pre-training of 3D molecular representations (MolSpectra), thereby infusing the knowledge of quantum mechanics into the molecular representations. Specifically, we propose SpecFormer, a multi-spectrum encoder for encoding molecular spectra via masked patch reconstruction. By further aligning outputs from the 3D encoder and spectrum encoder using a contrastive objective, we enhance the 3D encoder's understanding of molecules. Evaluations on public benchmarks reveal that our pre-trained representations surpass existing methods in predicting molecular properties and modeling dynamics.