Molecular Inverse-Design Platform for Material Industries

• DOI: 10.1145/3394486.3403346
• Publication Date: 2020-08-20T23:03:55Z
• Authors: Seiji Takeda; Toshiyuki Hama; Hsiang-Han Hsu; Victoria A. Piunova; Dmitry Zubarev; Daniel P. Sanders; Jed W. Pitera; Makoto Kogoh; Takumi Hongo; Yenwei Cheng; Wolf Bocanett; Hideaki Nakashika; Akihiro Fujita; Yuta Tsuchiya; Katsuhiko Hino; Kentaro Yano; Shuichi Hirose; Hiroki Toda; Yasumitsu Orii; Daiju Nakano
• Keywords: FOS: Computer and information sciences; FOS: Physical sciences; 02 engineering and technology; 01 natural sciences; 0104 chemical sciences; Computational Engineering, Finance, and Science (cs.CE); Physics - Data Analysis, Statistics and Probability; Computer Science - Computational Engineering, Finance, and Science; 0210 nano-technology; Data Analysis, Statistics and Probability (physics.data-an)

The discovery of new materials has been the essential force which brings a discontinuous improvement to industrial products' performance. However, extra-vast combinatorial design space material structures exceeds human experts' capability explore all, thereby hampering development. In this paper, we present industry-oriented web platform an AI-driven molecular inverse-design system, automatically designs brand rapidly and diversely. Different from existing solutions, in combination substructure-based feature encoding graph generation algorithms allows user gain high-speed, interpretable, customizable process. Also, hierarchical data structure user-oriented UI provide flexible intuitive workflow. system is deployed on IBM's our client's cloud servers used by 5 partner companies. To illustrate actual use cases, exhibit sugar dye molecules, that were carried out experimental chemists those client Compared general chemist's standard performance, speed was accelerated more than 10 times, greatly increased variety observed inverse-designed molecules without loss chemical realism.