FRIGID: Scaling Diffusion-Based Molecular Generation from Mass Spectra at Training and Inference Time

2026-04-17

AI Summary

This work addresses the challenging problem of de novo molecular structure elucidation from mass spectra by proposing FRIGID, a diffusion-based language modeling framework. Pretrained on hundreds of millions of unlabeled molecules, FRIGID generates candidate structures conditioned on input mass spectra through an intermediate fingerprint representation and molecular formula constraints. A forward fragmentation model is further integrated during inference to correct structural fragments inconsistent with the observed spectrum. The approach innovatively extends both training and inference phases: leveraging massive unlabeled data during training and employing targeted remasking and denoising strategies at inference time to dynamically refine predictions, yielding log-linear performance gains with increased computational resources. On the MassSpecGym and NPLIB1 benchmarks, FRIGID achieves Top-1 accuracy exceeding 18% and outperforms prior state-of-the-art methods by a factor of three, substantially advancing the performance frontier in this domain.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

In this work, we present FRIGID, a framework with a novel diffusion language model that generates molecular structures conditioned on mass spectra via intermediate fingerprint representations and determined chemical formulae, training at the scale of hundreds of millions of unlabeled structures. We then demonstrate how forward fragmentation models enable inference-time scaling by identifying spectrum-inconsistent fragments and refining them through targeted remasking and denoising. While FRIGID already achieves strong performance with its diffusion base, inference-time scaling significantly improves its accuracy, surpassing 18% Top-1 accuracy on the challenging MassSpecGym benchmark and tripling the Top-1 accuracy of the leading methods on NPLIB1. Further empirical analyses show that FRIGID exhibits log-linear performance scaling with increasing inference-time compute, opening a promising new direction for continued improvements in de novo structural elucidation. FRIGID code is publicly available at https://github.com/coleygroup/FRIGID

Problem

• molecular generation
• mass spectra
• structural elucidation
• de novo
• diffusion model

Innovation

• diffusion language model
• inference-time scaling
• mass spectra
• molecular generation
• fragmentation modeling