One Small Step with Fingerprints, One Giant Leap for De Novo Molecule Generation from Mass Spectra

Neng Kai Nigel Neo, Jing Lim, Ngoui Yong Zhau Preston, Koh Xue Ting Serene, Bingquan Shen

Keywords: De Novo Molecule Generation, Mass Spectrometry, Automated Characterization

TL;DR: We use MIST to encode mass spectra to fingerprints and MolForge to decode the fingerprints to structure to achieve top-1 28% / top-10 36% accuracy from mass spectra in MassSpecGym, a tenfold improvement over state-of-the-art.

Abstract: A common approach to the de novo molecular generation problem from mass spectra involves a two-stage pipeline: (1) encoding mass spectra into molecular fingerprints, followed by (2) decoding these fingerprints into molecular structures. In our work, we adopt MIST (Goldman et. al., 2023) as the encoder and MolForge (Ucak et. al., 2023) as the decoder, leveraging additional training data to enhance performance. We also threshold the probabilities of each fingerprint bit to focus on the presence of substructures. This results in a tenfold improvement over previous state-of-the-art methods, generating top-1 31% / top-10 40% of molecular structures correctly from mass spectra in MassSpecGym (Bushuiev et. al., 2024). We position this as a strong baseline for future research in de novo molecule elucidation from mass spectra.

Submission Track: Paper Track (Short Paper)

Submission Category: Automated Material Characterization

Institution Location: {Singapore, Singapore}

Submission Number: 27