Generative pseudo-force fields for molecular generation

Abstract Generating stable molecular conformations typically forces a tradeoff between the physical realism of energy-based relaxation and the sampling efficiency of data-driven generative models. While machine learning force fields (MLFFs) can sample stable conformations by relaxing molecular geometries according to physical forces, they require costly ab-initio training data. Conversely, diffusion models (DMs) learn from equilibrium data alone but are dependent on noise schedules and time-step conditioning. In this work, we propose generative pseudo-force fields (GPFFs) to bridge these paradigms by training an MLFF on a quadratic pseudo-potential energy surface relative to reference equilibrium structures. Because no ab-initio calculations are required for the perturbed geometries, non-equilibrium training data can be generated on the fly by perturbing the equilibria with Gaussian noise. We show that GPFFs constitute a time-step-agnostic variant of variance-exploding DMs: the score comes from the predicted pseudo-forces but because force magnitudes implicitly encode the noise level, no time-step conditioning is needed. Our GPFF can hence be used as a drop-in replacement in standard diffusion sampling (ancestral, Heun) but also facilitates more efficient, adaptive variants and an MLFF inspired direct denoising scheme. Our proposed sampling algorithms support arbitrary structural priors and geometric constraints. On QM9, GPFF reaches 99 % validity with stochastic direct denoising at 256 neural function evaluations (NFE), and about 90 % at just 16 NFE with direct denoising, outperforming a matched diffusion baseline that shares its backbone, training recipe, and noise schedule across all samplers. Combined with custom priors, we showcase the fast and accurate generation process of our method in a molecular editor for a drug design setting.

Authors

Institutions

Publication Details

Journal
Machine Learning Science and Technology
Published
2026-11-16
DOI
https://doi.org/10.1088/2632-2153/aec2a2
Citations
3
Primary Topic
Computational Drug Discovery Methods
Type
article
Field-Weighted Citation Impact
3.97

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Generative pseudo-force fields for molecular generation

Stefan Gugler, Niklas W. A. Gebauer, Stefaan S. P. Hessmann, Michael Plainer et al.
3 citations
Machine Learning Science and Technology
Computational Drug Discovery Methods
3.97
article

Generative pseudo-force fields for molecular generation

Stefan Gugler, Niklas W. A. Gebauer, Stefaan S. P. Hessmann, Michael Plainer, Khaled Kahouli, Klaus Robert Müller, Frank Noé
article en
3 citations

Abstract

Abstract Generating stable molecular conformations typically forces a tradeoff between the physical realism of energy-based relaxation and the sampling efficiency of data-driven generative models. While machine learning force fields (MLFFs) can sample stable conformations by relaxing molecular geometries according to physical forces, they require costly ab-initio training data. Conversely, diffusion models (DMs) learn from equilibrium data alone but are dependent on noise schedules and time-step conditioning. In this work, we propose generative pseudo-force fields (GPFFs) to bridge these paradigms by training an MLFF on a quadratic pseudo-potential energy surface relative to reference equilibrium structures. Because no ab-initio calculations are required for the perturbed geometries, non-equilibrium training data can be generated on the fly by perturbing the equilibria with Gaussian noise. We show that GPFFs constitute a time-step-agnostic variant of variance-exploding DMs: the score comes from the predicted pseudo-forces but because force magnitudes implicitly encode the noise level, no time-step conditioning is needed. Our GPFF can hence be used as a drop-in replacement in standard diffusion sampling (ancestral, Heun) but also facilitates more efficient, adaptive variants and an MLFF inspired direct denoising scheme. Our proposed sampling algorithms support arbitrary structural priors and geometric constraints. On QM9, GPFF reaches 99 % validity with stochastic direct denoising at 256 neural function evaluations (NFE), and about 90 % at just 16 NFE with direct denoising, outperforming a matched diffusion baseline that shares its backbone, training recipe, and noise schedule across all samplers. Combined with custom priors, we showcase the fast and accurate generation process of our method in a molecular editor for a drug design setting.

Machine Learning Science and Technology
Technische Universität Berlin (DE), Freie Universität Berlin (DE)
National Science Foundation, Deutscher Akademischer Austauschdienst, Deutsche Forschungsgemeinschaft, Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung, Bundesministerium für Bildung und Forschung, Korea University, Ministry of Science and ICT, South Korea, Institute for Information and Communications Technology Promotion
Openalex Percentile: Top 7%
Computational Drug Discovery Methods
3.97
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon