Latent Diffusion-based 3D Molecular Recovery from Vibrational Spectra

Wenjin Wu, Ales Leonardis, Linjiang Chen, Jianbo Jiao

Keywords: Diffusion Models, Molecular Generation, Infrared Spectroscopy

Abstract:

Infrared (IR) spectroscopy, a type of vibrational spectroscopy, is widely used for molecular structure determination and provides critical structural information for chemists.
However, existing approaches for recovering molecular structures from IR spectra typically rely on one-dimensional SMILES strings or two-dimensional molecular graphs, which fail to capture the intricate relationship between spectral features and three-dimensional molecular geometry.
Recent advances in diffusion models have greatly enhanced the ability to generate molecular structures in 3D space.
Yet, no existing model has explored the distribution of 3D molecular geometries corresponding to a single IR spectrum.
In this work, we introduce *IR-GeoDiff*, a latent diffusion model that integrates IR spectral information into both node and edge representations of molecular structures. We evaluate IR-GeoDiff from both spectral and structural perspectives, demonstrating its ability to recover the molecular distribution corresponding to a given IR spectrum.
Furthermore, an attention-based analysis reveals that the model’s interpretation of IR spectra aligns with quantum mechanical principles of molecular vibrations.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Submission Number: 5905