# Unlocking High-Fidelity Molecular Generation from Mass Spectra via Dual-Stream Line Graph Diffusion

arXiv: https://arxiv.org/abs/2605.07048 · HTML: https://mlsanity.com/paper/2605.07048/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 42 references that are in the map.
- First posted on arXiv 7 May 2026.

## Builds on (most cited first, 18 of 18)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [FiLM: Visual Reasoning with a General Conditioning Layer](https://mlsanity.com/paper/1709.07871/) (2017)
- [Structured Denoising Diffusion Models in Discrete State-Spaces](https://mlsanity.com/paper/2107.03006/) (2021)
- [Rethinking Attention with Performers](https://mlsanity.com/paper/2009.14794/) (2020)
- [Directional Message Passing for Molecular Graphs](https://mlsanity.com/paper/2003.03123/) (2020)
- [DiGress: Discrete Denoising diffusion for graph generation](https://mlsanity.com/paper/2209.14734/) (2022)
- [GemNet: Universal Directional Graph Neural Networks for Molecules](https://mlsanity.com/paper/2106.08903/) (2021)
- [Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models](https://mlsanity.com/paper/1811.12823/) (2018)
- [ChemRL-GEM: Geometry Enhanced Molecular Representation Learning for Property Prediction](https://mlsanity.com/paper/2106.06130/) (2021)
- [MassSpecGym: A benchmark for the discovery and identification of molecules](https://mlsanity.com/paper/2410.23326/) (2024)
- [DiffMS: Diffusion Generation of Molecules Conditioned on Mass Spectra](https://mlsanity.com/paper/2502.09571/) (2025)
- [Fast Sampling via Discrete Non-Markov Diffusion Models with Predetermined Transition Time](https://mlsanity.com/paper/2312.09193/) (2023)
- [MADGEN: Mass-Spec attends to De Novo Molecular generation](https://mlsanity.com/paper/2501.01950/) (2025)
- [Triplet Interaction Improves Graph Transformers: Accurate Molecular Graph Learning with Triplet Graph Transformers](https://mlsanity.com/paper/2402.04538/) (2024)
- [Competitive Fragmentation Modeling of ESI-MS/MS spectra for putative metabolite identification](https://mlsanity.com/paper/1312.0264/) (2013)
- [MS-BART: Unified Modeling of Mass Spectra and Molecules for Structure Elucidation](https://mlsanity.com/paper/2510.20615/) (2025)
- [Comparative Analysis of Formula and Structure Prediction from Tandem Mass Spectra](https://mlsanity.com/paper/2601.00941/) (2026)
- [De Novo Molecular Generation from Mass Spectra via Many-Body Enhanced Diffusion](https://mlsanity.com/paper/2602.01643/) (2026)

## Abstract (arXiv)

De novo molecular generation from tandem mass spectra is a challenging inverse problem whose core difficulty lies in the circular dependency between atom-level and bond-level reasoning: determining a bond's type requires knowing its endpoint atoms' chemical environment, yet an atom's environment is in turn defined by its incident bonds. Existing graph diffusion methods process atoms and bonds within a single computation stream, where atom-bond information synchronization can only occur implicitly across layers. We argue that this single-stream paradigm, rather than the choice of any particular aggregation kernel, is a key architectural bottleneck. We propose DualLGD (Dual-stream Line Graph Diffusion), which reformulates molecular graph denoising as the alternating solution of two coupled subproblems: atom-level reasoning and bond-level reasoning, each operating in its own dedicated representation space. The line graph provides a natural mathematical construction for the bond space, in which bond angles, dihedrals, conjugation chains, and rings correspond to local topological motifs between bonds. Incidence-constrained bidirectional cross-attention synchronizes the two streams at every layer, ensuring that each atom attends only to its incident bonds and vice versa, respecting the fundamental chemical principle that an atom's environment is determined by its bonding context. On the NPLIB1 and MassSpecGym benchmarks, DualLGD achieves top-1 accuracy of 34.37% and 23.89%, approximately 3× the previous state of the art. Ablation studies confirm the architecture as the primary source of improvement: DualLGD without any pre-training already surpasses the previous best fully pretrained model.