NMIRacle: Multi-modal Generative Molecular Elucidation from IR and NMR Spectra

TL;DR

NMIRacle addresses de novo molecular structure elucidation from raw IR and NMR spectra by proposing a two-stage conditional generative framework. A count-aware fragment prior built from a 991-pattern fragment vocabulary ($p_\phi(\mathbf{y}|\mathbf{c})$) is learned in Stage 1, while a multi-spectral encoder ($q_\psi$) maps spectra to a latent conditioning variable that guides Stage 2 generation. The combination of a fragment-centric representation with a dedicated multimodal encoder enables direct spectra-to-molecule generation and shows state-of-the-art results on unseen molecules, with robust scaling to more complex structures. By reducing heavy pre-processing and leveraging multi-modal spectral evidence, NMIRacle advances realistic, data-driven structure elucidation from spectroscopic data.

Abstract

Molecular structure elucidation from spectroscopic data is a long-standing challenge in Chemistry, traditionally requiring expert interpretation. We introduce NMIRacle, a two-stage generative framework that builds upon recent paradigms in AI-driven spectroscopy with minimal assumptions. In the first stage, NMIRacle learns to reconstruct molecular structures from count-aware fragment encodings, which capture both fragment identities and their occurrences. In the second stage, a spectral encoder maps input spectroscopic measurements (IR, 1H-NMR, 13C-NMR) into a latent embedding that conditions the pre-trained generator. This formulation bridges fragment-level chemical modeling with spectral evidence, yielding accurate molecular predictions. Empirical results show that NMIRacle outperforms existing baselines on molecular elucidation, while maintaining robust performance across increasing levels of molecular complexity.

Figures (8)

• Figure 1: Overview of the NMIRacle framework. (a) The model is trained in two stages: Stage 1 learns a fragment-conditioned molecular generator that reconstructs full molecular structures from count-aware fragment representations, establishing a molecular prior $p_\phi(\mathbf{y}\mid\mathbf{c})$. Stage 2 introduces a multi-spectral encoder that maps raw IR, ${}^1$H-NMR, and ${}^{13}$C-NMR spectra into latent embeddings $\mathbf{z}_\psi(\mathcal{S})$, used to condition the pre-trained generator for direct spectra-to-molecule generation. (b) The architecture integrates (i) a count-aware fragment encoder that embeds molecular fragments and their occurrences, and (ii) a multi-spectral encoder that fuses complementary spectral modalities into a shared latent representation.
• Figure 2: (Top) Venn diagrams illustrate the overlap between the molecular pre-training dataset (derived from GDB-17 and SpectraBase) and the additional molecules incorporated from alberts2024unraveling dataset. (Bottom) element distribution across the utilized datasets, highlighting the broader chemical diversity introduced by the data augmentation.
• Figure 3: Model performance across molecular complexity bins. NMIRacle maintains higher Tanimoto similarity even for structurally-rich molecules.
• Figure 4: (Top) Mean F1-score across fragments grouped by molecular frequency. (Bottom) Cumulative coverage of the top-$X\%$ fragments ranked by F1-score (for the IR + ${}^{1}$H-NMR + ${}^{13}$C-NMR setting), plotted against their total occurrence fraction. Together, the plots show that accuracy increases for frequent motifs, with a subset of high-scoring fragments covering most observed cases.
• Figure 5: Ablation results comparing the impact of learnable positional encodings (left) and inter-modal transformer encoder (right). Bars report relative performance with respect to the full NMIRacle configuration (blue).