Symmetry-Aware Flow Matching for End-to-end Molecular Crystal Generation

英文原文与中文译文逐段对照 · English original with Chinese translation

Wendi Cai${}^{ \dagger }$ and Fanyang Mo${}^{ * }$

蔡文迪${}^{ \dagger }$ 和莫凡洋${}^{ * }$

${}^{1}$ School of AI for Science, Peking University Shenzhen Graduate School, Shenzhen 518055, China; ${}^{2}$ School of Advanced Materials, Peking University Shenzhen Graduate School, Shenzhen 518055, China; ${}^{3}$ Guangdong Provincial Key Laboratory of Nano-Micro Materials Research, Peking University Shenzhen Graduate School, Shenzhen 518055, China; ${}^{4}$ State Key Laboratory of Advanced Waterproof Materials, School of Materials Science and Engineering, Peking University, Beijing 100871, China. ${}^{ \dagger }$ Work completed during an internship at ByteDance. ${}^{ * }$ Corresponding author.

${}^{1}$ 北京大学深圳研究生院科学智能学院；${}^{2}$ 北京大学深圳研究生院新材料学院；${}^{3}$ 广东省纳米微米材料研究重点实验室；${}^{4}$ 北京大学材料科学与工程学院先进防水材料国家重点实验室。${}^{ \dagger }$ 主要工作于字节跳动实习期间完成。${}^{ * }$ 通信作者.

Cite This: arXiv: 2609.34690 (2026). https://arxiv.org/abs/2609.34690. Submitted September 28, 2026.

引用本文: arXiv: 2609.34690 (2026). https://arxiv.org/abs/2609.34690. 2026年9月28日提交.

ABSTRACT: Molecular crystal packing shapes properties from drug bioavailability to charge transport, yet generating realistic structures requires coordinating molecular flexibility, intermolecular interactions and symmetry. Here we introduce SALA, a flow-matching model built on state–context separation: it evolves only asymmetric-unit (ASU) atomic coordinates and symmetry-compatible lattice parameters while predicting their updates from the full periodic environment. Conditioned on molecular graphs and a space group, SALA generates molecular conformation, lattice geometry and packing end to end while preserving the specified symmetry. Across 799 held-out crystals spanning four chemical classes and all seven crystal systems, SALA achieves a 72.0% hit rate from 50 candidates per target, compared with 8.1% for an atomistic full-cell generative baseline. Mean best-candidate packing root-mean-square deviation decreases from 4.48 to 1.70 Å, with the advantage widening as system size and symmetry multiplicity increase. SALA thus enables end-to-end generation of complex, symmetry-constrained molecular crystals, with potential applications in CSP, structure-based pharmaceutical polymorph design and organic molecular materials discovery.

摘要：分子晶体的堆积方式决定了从药物生物利用度到电荷传输等一系列性质，然而生成真实合理的结构需要协同处理分子柔性、分子间相互作用与对称性。在此，我们提出 SALA——一个建立在"状态—上下文分离"（state–context separation）之上的流匹配（flow matching）模型：它仅演化不对称单元（asymmetric unit, ASU）的原子坐标以及与对称性相容的晶格参数，同时从完整的周期性环境中预测这些量的更新。以分子图和空间群为条件，SALA 能够端到端地生成分子构象、晶格几何与堆积方式，并保持指定的对称性。在涵盖四类化合物与全部七个晶系的 799 个留出晶体上，SALA 以每个目标 50 个候选达到 72.0% 的命中率，而基于原子级全晶胞生成的基线仅为 8.1%。最优候选的平均堆积均方根偏差（packing RMSD）由 4.48 Å 降至 1.70 Å，且随着体系尺寸与对称多重度的增加，优势进一步扩大。因此，SALA 实现了复杂、受对称性约束的分子晶体的端到端生成，在晶体结构预测（CSP）、基于结构的药物多晶型设计以及有机分子材料发现等方面具有应用潜力。

注：本文为 arXiv 预印本，全文双语正文待补充。预印本全文：arXiv: 2609.34690。