Mitsui & Co's Xeureka Utilizes NVIDIA AI for Secure Drug Discovery

In an innovative stride towards advancing healthcare research, Xeureka, a subsidiary of Mitsui & Co, is leveraging NVIDIA's AI platform to enhance drug discovery through secure data sharing, according to a report by [source name](https://blogs.nvidia.com/blog/mitsui-xeureka-drug-discovery/). This initiative is part of Mitsui & Co's broader digital transformation strategy, which integrates cutting-edge technologies such as AI and confidential computing.

Digital Transformation and Xeureka's Mission

Established 77 years ago, Mitsui & Co, a Tokyo-based conglomerate, is embracing digital transformation across its 16 divisions. Among its ventures is Xeureka, a subsidiary focused on accelerating research and development in healthcare. The process of bringing a new drug to market can be prohibitively expensive, often requiring over a billion dollars and a decade of work. Xeureka aims to mitigate these challenges by utilizing AI and confidential computing to facilitate the safe sharing of critical data.

Confidential Computing: A Solution to Data Sharing

Xeureka's approach involves confidential computing, a technology that processes data in a secure environment within a GPU or CPU. This method ensures that sensitive information, such as patient data and proprietary chemical formulas, remains protected, enabling pharmaceutical companies to share datasets without compromising privacy. By collaborating with tech companies like NVIDIA and Fortanix, Xeureka has developed a proof of concept that leverages this technology to address data scarcity in drug discovery.

Proof of Concept and Enhanced Accuracy

In a demonstration of confidential computing's potential, Xeureka conducted a proof of concept involving two fictional companies, each possessing datasets of a thousand drug candidates. These datasets were used to train AI models to predict chemical toxicity. When combined, the datasets allowed for the creation of a larger, more accurate model, achieving prediction accuracy improvements of 65-74%. This success underscores the technology's ability to enhance model stability and reduce bias, as highlighted by Xeureka's CTO, Hiroki Makiguchi.

Future Prospects with Tokyo-1 Supercomputer

Looking forward, Xeureka plans to expand its application of AI in drug discovery through Tokyo-1, a GPU-accelerated AI supercomputer. Announced in February, Tokyo-1 is designed to boost the efficiency of pharmaceutical research in Japan and beyond. Potential projects include predicting protein structures and accelerating molecular dynamics simulations. Through collaborations with the Tokyo-1 community, Xeureka aims to harness large language models for chemistry and genetic data, furthering its mission to transform the pharmaceutical industry.

This initiative aligns with Mitsui's strategic growth plan to enhance software and services for healthcare, particularly in Japan's $100 billion pharmaceutical market, the third largest globally. Xeureka's services promise to revolutionize the industry by enabling rapid screening of drug candidates and simulating chemical interactions.