JSR Corporation (Representative Director, President & CEO: Tetsuro Hori, hereinafter "JSR") and IBM today announced that a joint research outcome from their collaborative AI research program for the chemical industry has been accepted for publication in Nature Computational Science. The collaboration is being conducted under a Joint Research Agreement (JRA) between the two organizations. Nature Computational Science, published by Nature Portfolio, is a leading international scientific journal featuring research that applies computational science to address complex challenges across natural sciences, engineering, and related disciplines.

The published research combines IBM's expertise in AI and algorithms with JSR's deep knowledge of chemistry and materials science to achieve both high predictive accuracy and interpretability in the prediction of material properties for organic materials.

The method developed through this research leverages topological distance information within molecules, enabling it to capture complex structural characteristics of entire molecules that are difficult to represent using conventional approaches. Researchers from both organizations validated the predictions generated by the method through molecular orbital analysis. The results demonstrated superior performance compared with representative existing methods in predicting the optical properties of complex dye molecules, while also providing chemically interpretable explanations for the predictions. The research findings offer new insights for foundation models that handle complex molecular structures as well as for quantum chemistry calculations. Looking ahead, the approach is expected to contribute to more efficient materials development and the discovery of promising material candidates through applications in materials screening and molecular design.

Takeshi Fukuda, Executive Officer, Head of Research & Development, and Director of IBM Research - Tokyo, IBM Japan, said: "This achievement marks an important step forward in Material DX, where AI and materials experts work together to accelerate the discovery of new materials. Through our continued collaboration with JSR, we are committed to advancing materials development technologies by leveraging AI and quantum computing."

Toru Kimura, Senior Executive Officer, Head of Electronic Materials Business, Next Generation Research (DS/MI, etc.), JSR Corporation, said: "This achievement is significant, demonstrating the unprecedented potential of integrating AI and materials science in materials development. JSR will further enhance its research and development process, from materials exploration and design to experimentation and verification, by combining its long-cultivated materials expertise with cutting-edge digital technologies, including AI. Moving forward, through advanced initiatives and partnership with IBM, we aim to dramatically increase the speed and quality of development to deliver materials with new value to society more quickly."

As emerging technologies and digital transformation continue to accelerate, JSR and IBM are working together to help lead a new era of technological and societal advancement. JSR's materials innovation in the semiconductor field forms part of the backbone of a wide range of industries, helping enable a more connected world through technologies that are seamlessly integrated into everyday life.

Title: Revisiting Molecular Descriptors with TDiMS for Interpretable Intramolecular Interactions Based on Substructure Pairs

Authors: Lisa Hamada¹, Akihiro Kishimoto¹, Kohei Miyaguchi¹, Yoshitaka Hirose², Junta Fuchiwaki², Indra Priyadarsini¹, and Seiji Takeda¹

¹ IBM Research ² JSR Corporation

Publication URL: https://www.nature.com/articles/s43588-026-01036-3