User-Interactive Molecular Graph Suggestion for Drug Discovery

• Item type: SIG Technical Reports(1)
• 公開日: 2022-09-02
• タイトル: User-Interactive Molecular Graph Suggestion for Drug Discovery
• タイトル / 言語: en
• タイトル: User-Interactive Molecular Graph Suggestion for Drug Discovery
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: Hokkaido University
• 著者所属: Hokkaido University／RIKEN AIP
• 著者所属: Nagoya University／Osaka University
• 著者所属(英): en
• 著者所属(英): Hokkaido University
• 著者所属(英): en
• 著者所属(英): Hokkaido University / RIKEN AIP
• 著者所属(英): en
• 著者所属(英): Nagoya University / Osaka University
• 著者名: Sheng, Hu; Ichigaku, Takigawa; Chuan, Xiao
• 著者名(英): Sheng, Hu; Ichigaku, Takigawa; Chuan, Xiao
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: We present a novel molecular graph generation method by auto-completing a privileged scaffold which represents a core graph substructure step-by-step. We propose a generative GNN model thus providing the ability to generate unseen molecular graphs outside the given training set. An edit-aware graph autocompletion paradigm that follows the “substructure-by-substructure” process is designed to complete the scaffold queries in multiple substructure adopt operations and allow meaningful edit operations to show the user's intention. Such operations enable the involvement of user decisions when interacting with a generative user-centered AI system, which differentiates our work from existing single-run generation paradigms. Particularly, a Monte Carlo tree search (MCTS) method is employed to satisfy the property requirements and navigate the search space when complying with users' edit operations. Moreover, we design a top-k ranking function which considers the preferences on popularity and diversity for different applications, such as query compositions for graph database and drug discovery respectively. Such techniques enable human experts to synergistically interact with the generative models grounded on large data.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: We present a novel molecular graph generation method by auto-completing a privileged scaffold which represents a core graph substructure step-by-step. We propose a generative GNN model thus providing the ability to generate unseen molecular graphs outside the given training set. An edit-aware graph autocompletion paradigm that follows the “substructure-by-substructure” process is designed to complete the scaffold queries in multiple substructure adopt operations and allow meaningful edit operations to show the user's intention. Such operations enable the involvement of user decisions when interacting with a generative user-centered AI system, which differentiates our work from existing single-run generation paradigms. Particularly, a Monte Carlo tree search (MCTS) method is employed to satisfy the property requirements and navigate the search space when complying with users' edit operations. Moreover, we design a top-k ranking function which considers the preferences on popularity and diversity for different applications, such as query compositions for graph database and drug discovery respectively. Such techniques enable human experts to synergistically interact with the generative models grounded on large data.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AN10114171
• 書誌情報: 研究報告情報基礎とアクセス技術（IFAT）; 巻 2022-IFAT-148, 号 33, p. 1-5, 発行日 2022-09-02
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8884
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会