[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-117371-en":3,"doc-seo-117371-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},117371,687197207639,"Asher","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks - Research paper","A molecule’s 2D representation encodes atoms, their attributes, and covalent connectivity, while a 3D conformer representation includes atom types and Cartesian coordinates with associated potential energy. Lower-energy conformers occur more often in nature, yet most molecular property prediction models treat either 2D graphs or single 3D conformers in isolation. The proposed E(3)-invariant aggregation network unifies 2D features with multiple conformers via a differentiable Fused Gromov-Wasserstein barycenter solver and efficient distance-geometry conformer generation. An optimized GPU implementation and experiments show substantial gains over state-of-the-art methods on established datasets.","Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks  \nDuy M. H. Nguyen * 1 2 3 Nina Lukashina * 1 2 Tai Nguyen 3 An T. Le 4 TrungTin Nguyen 5 Nhat Ho 6  \nJan Peters 3 4 7 Daniel Sonntag 3 8 Viktor Zaverkin 9 Mathias Niepert 1 2 9  \nAbstract  \nA molecule’s 2D representation consists of its atoms, their attributes, and the molecule’s covalent bonds. A 3D (geometric) representation of a molecule is called a conformer and consists of its atom types and Cartesian coordinates. Every conformer has a potential energy, and the lower this energy, the more likely it occurs in nature. Most existing machine learning methods for molecular property prediction consider either 2D molecular graphs or 3D conformer structure representationsin isolation. Inspired by recent work on using ensembles of conformers in conjunction with 2D graph representations, we propose E(3)-invariant molecular conformer aggregation networks. The method integrates a molecule’s 2D representation with that of multiple of its conformers. Contrary to prior work, we propose a novel 2D–3D aggregation mechanism based on a differentiable solver for the Fused Gromov-Wasserstein Barycenter problem and the use of an efficient conformer generation method based on distance geometry.  \nWe show that the proposed aggregation mechanism is E(3) invariant and propose an efficient GPU implementation. Moreover, we demonstrate that the aggregation mechanism helps to significantly outperform state-of-the-art molecule property prediction methods on established datasets.  \nOur implementation is available at this link.  \n1. Introduction  \nMachine learning is increasingly used for modeling and analyzing properties of atomic systems with important ap-  \n*Equal contribution 1Department of Computer Science, University of Stuttgart, Germany 2Max Planck Research School for Intelligent Systems (IMPRS-IS) 3 German Research Center for Artificial Intelligence (DFKI) 4Department of Computer Science, Technische Universitat Darmstadt, Germany 5 School of Mathematics and Physics, University of Queensland, Australia 6Department of Statistics and Data Sciences, University of Texas at Austin, USA 7Hessian.AI 8Department of Applied Artificial Intelligence, Oldenburg University, Germany 9NEC Laboratories Europe. Correspondence to: Duy H. M. Nguyen \u003C[hong01@dfki.de](hong01@dfki.de) > .  \nProceedings of the 41 st International Conference on Machine Learning, Vienna, Austria. PMLR 235, 2024 . Copyright 2024 by the author(s) .  \nplications in drug discovery and material design (Butler et al., 2018 ; Vamathevan et al., 2019 ; Choudhary et al., 2022 ; Fedik et al., 2022 ; Batatia et al., 2023) . Most existing machine learning approaches to molecular property prediction either incorporate 2D (topological) (Kipf & Welling, 2017 ; Gilmer et al., 2017b ; Xu et al., 2018 ; Velikoviet al., 2018) or 3D (geometric) information of molecular structures (Sch¨utt et al., 2017 ; Sch¨utt et al., 2021 ; Batzner et al., 2022 ; Batatia et al., 2022) . 2D molecular graphs describe molecular connectivity (covalent bonds) but ignore the spatial arrangement of the atoms in a molecule (molecular conformation) . 3D graph representations capture conformational changes but are commonly used to encode an individual conformer. Many molecular properties, such as solubility and binding affinity (Cao et al., 2022), however, inherently depend on a large number of conformations a molecule can occur as in nature, and employing a single geometry per molecule limits the applicability of machine-learning models. Furthermore, it is challenging to determine conformers that predominantly contribute to the molecular properties of interest. Thus, developing expressive representations for molecular systems when modeling their properties is an ongoing challenge.  \nTo overcome this, recent work has introduced molecular representations that incorporate both 2D molecular graphs and 3D conformers (Zhu et al., 2023) . These methods aim to encode","cbCaierGLixtfo0Q","https://ap.wps.com/l/cbCaierGLixtfo0Q","pdf",2868551,1,25,"English","en",105,"# Introduction\n## Problem background: 2D vs 3D molecular representations\n## Proposed contributions\n### E(3)-invariant 2D–3D conformer aggregation\n### Differentiable FGW barycenter solver and GPU acceleration\n### Conformer sampling via distance geometry","[{\"question\":\"What limitation do existing molecular ML methods face when using 2D or 3D inputs?\",\"answer\":\"Many approaches either use 2D molecular graphs or 3D conformer representations separately, and single-geometry models fail to reflect the many conformations a molecule can adopt in nature.\"},{\"question\":\"How does the proposed method combine 2D molecular graphs with multiple 3D conformers?\",\"answer\":\"It introduces an E(3)-invariant conformer ensemble aggregation mechanism that integrates 2D representations with multiple conformers using a differentiable Fused Gromov-Wasserstein barycenter approach.\"},{\"question\":\"Why is the aggregation mechanism suitable for large-scale training?\",\"answer\":\"The work accelerates the FGW barycenter solver using entropic-based techniques, enabling parallel training across multiple GPUs.\"}]","Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks - Research paper | PDF",1785675430,63,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"structure-aware-e3-invariant-molecular-conformer-aggregation-networks-research-paper","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/structure-aware-e3-invariant-molecular-conformer-aggregation-networks-research-paper/117371/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-02",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What limitation do existing molecular ML methods face when using 2D or 3D inputs?","Question",{"text":75,"@type":76},"Many approaches either use 2D molecular graphs or 3D conformer representations separately, and single-geometry models fail to reflect the many conformations a molecule can adopt in nature.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed method combine 2D molecular graphs with multiple 3D conformers?",{"text":80,"@type":76},"It introduces an E(3)-invariant conformer ensemble aggregation mechanism that integrates 2D representations with multiple conformers using a differentiable Fused Gromov-Wasserstein barycenter approach.",{"name":82,"@type":73,"acceptedAnswer":83},"Why is the aggregation mechanism suitable for large-scale training?",{"text":84,"@type":76},"The work accelerates the FGW barycenter solver using entropic-based techniques, enabling parallel training across multiple GPUs.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]