arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.07693cs.LG
收藏

迈向更优的分子生成模型几何表示

Toward Better Geometric Representations for Molecule Generative Models

Institute for Artificial Intelligence, PKU(北京大学人工智能学院);ByteDance AI Drug Discovery(字节跳动AI药物发现);School of Intelligence Science and Technology, PKU(北京大学智能科学与技术学院);Anew Labs(Anew实验室);Department of Electrical Engineering and Computer Science, MIT(麻省理工学院电子工程与计算机科学系);State Key Laboratory of General Artificial Intelligence, PKU(北京大学通用人工智能国家重点实验室)

Shaoheng Yan, Zian Li, Cai Zhou, Qiaojing Huang, Kai Liu, Muhan Zhang

AI总结:

本文提出LENSEs框架,通过改进分子表示方法提升生成质量,实验显示其在GEOM-DRUG数据集上取得高有效性和稳定性。

摘要:

几何表示条件下的分子生成提供了一种有效范式,将分子表示建模与结构生成解耦。通过将分子生成分为两个阶段—首先生成有意义的分子表示,然后基于此表示生成3D分子—可以显著提高生成过程的效率和质量。然而,其有效性从根本上受限于表示空间的质量:预训练分子编码器,如UniMol,生成的表示非平滑且在生成训练过程中未被充分利用。在本文中,我们提出了LENSEs,一个框架,更有效地利用分子表示在表示条件生成方法中的潜力。特别是,LENSEs引入了三种互补机制:(1) 一个在生成任务中同时训练的表示头,从预训练编码器中提取多级表示;(2) 一个分子感知损失,优化生成器在语义信息丰富的表示空间中;(3) 一个节点级表示对齐(REPA)损失,明确对齐生成器的隐藏状态与编码器表示,减少预训练与生成之间的语义差距。我们通过广泛的分子生成任务验证了这些改进的有效性。具体而言,在具有挑战性的分子生成数据集GEOM-DRUG上,LENSEs实现了97.28%的有效性和98.51%的分子稳定性,超越现有先进方法。进一步通过Lipschitz常数减少(4.6倍)和QM9探测任务的分析也展示了更平滑、更信息丰富的细化表示,确立了具有对齐目标的生成训练作为分子编码器预训练范式的潜力。

英文摘要:

Geometric representation-conditioned molecule generation provides an effective paradigm that decouples molecule representation modeling from structure generation. By decoupling molecule generation into two stages-first generating a meaningful molecule representation, and then generating a 3D molecule conditioned on this representation-the efficiency and quality of the generation process can be significantly enhanced. However, its effectiveness is fundamentally limited by the quality of the representation space: pretrained molecular encoders, such as UniMol, produce representations that are non-smooth and not fully exploited during the generative training process. In this work, we propose LENSEs, a framework that better exploits the potential of molecule representations in representation-conditioned generation methods. In particular, LENSEs introduces three complementary mechanisms: (1) a representation head, simultaneously trained during generative tasks, that extracts multi-level representations from the pretrained encoder; (2) a molecule perceptual loss that optimizes the generator in a semantic-informative representation space; and (3) a node-level representation alignment (REPA) loss that explicitly aligns the generator's hidden states with encoder representations, reducing the semantic gap between pretraining and generation. We demonstrate the effectiveness of these improvements through extensive molecule generation tasks. Specifically, on the challenging molecule generation dataset GEOM-DRUG, LENSEs achieves 97.28% validity and 98.51% molecule stability, surpassing existing advanced methods. Further analyses through Lipschitz constant reduction (4.6x) and QM9 probing tasks also demonstrate the smoother, more informative refined representations, establishing generative training with alignment objectives as a potential pretraining paradigm for molecular encoders.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑