Fixed-Dimensional Latent Flow for Generating Variable-Size 3D Molecules
Abstract
Molecular size is intrinsically linked to composition, structure, and properties, but most 3D molecular generators treat size as a predefined input rather than a quantity to be generated. We introduce Equivariant-Free Transformer-Autoencoded Latent Flow Matching, a two-stage framework that samples a fixed-dimensional latent vector using flow matching and employs an autoregressive Transformer to determine molecular size, atom types, coordinates, and chemical attributes. Canonical atom ordering and rigid-pose alignment enable Transformers without equivariant layers, while decoded attributes guide bond reconstruction. On PCQM4Mv2, unconditional generation yields 87.9\% unique and novel molecules passing sanitization and PoseBusters checks, exceeding baselines with lower end-to-end training and sampling time and higher end-to-end throughput. Across ten target HOMO-LUMO gaps, internal ranking retains 30\% of the screened candidates and increases the density functional theory-verified hit rate within 0.1 eV of the target from 25.0\% to 52.4\%, while largely preserving novelty and diversity. These results demonstrate fixed-dimensional latent generation with autoregressive decoding as a practical approach to molecular design without prespecifying size.
Related
- A Latent Inflammatory Tissue-State Variable Mechanistically Links Radiotherapy-Induced Immune Remodeling to Recurrent Tumor Permissiveness
- RTMB: An R Package for Fitting Latent Variable Models With Template Model Builder
- Latent variable modelling of condition and reproductive investment in Gryllus vocalis field crickets
- From labels to latents: revealing state-dependent hippocampal computations and non-spatial replay with Jump Latent Variable Model
- Co-retained intron retention on single RNA molecules marks fitness-critical genes
BibTeX
@misc{yao2026fixeddimensional,
title = {{Fixed-Dimensional Latent Flow for Generating Variable-Size 3D Molecules}},
author = {Weichi Yao and Cameron Gruich and Bryan R. Goldsmith and Yixin Wang},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2609.08333},
eprint = {2609.08333},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.08333},
}
RIS
TY - UNPB
TI - Fixed-Dimensional Latent Flow for Generating Variable-Size 3D Molecules
AU - Weichi Yao
AU - Cameron Gruich
AU - Bryan R. Goldsmith
AU - Yixin Wang
PY - 2026
DA - 2026/10/21
PB - arXiv
DO - 10.48550/arxiv.2609.08333
UR - https://arxiv.org/abs/2609.08333
AB - Molecular size is intrinsically linked to composition, structure, and properties, but most 3D molecular generators treat size as a predefined input rather than a quantity to be generated. We introduce Equivariant-Free Transformer-Autoencoded Latent Flow Matching, a two-stage framework that samples a fixed-dimensional latent vector using flow matching and employs an autoregressive Transformer to determine molecular size, atom types, coordinates, and chemical attributes. Canonical atom ordering and rigid-pose alignment enable Transformers without equivariant layers, while decoded attributes guide bond reconstruction. On PCQM4Mv2, unconditional generation yields 87.9\% unique and novel molecules passing sanitization and PoseBusters checks, exceeding baselines with lower end-to-end training and sampling time and higher end-to-end throughput. Across ten target HOMO-LUMO gaps, internal ranking retains 30\% of the screened candidates and increases the density functional theory-verified hit rate within 0.1 eV of the target from 25.0\% to 52.4\%, while largely preserving novelty and diversity. These results demonstrate fixed-dimensional latent generation with autoregressive decoding as a practical approach to molecular design without prespecifying size.
ER -