The Tokyo-1 project announced today includes support for the NVIDIA DGX AI supercomputer for Japanese pharmaceutical companies and startups. This will accelerate Japan’s $100 billion pharmaceutical industry, which is the third largest in the world after the United States and China.
“Japanese pharmaceutical companies are experts in web lab research, but have yet to use high-performance computing and AI at scale,” said Yuhi Abe, head of digital health business unit at Mitsui. We are creating an innovation hub that can transform the landscape of the pharmaceutical industry with cutting-edge tools to accelerate drug discovery.
The project will give customers access to the NVIDIA DGX H100 node, which supports molecular dynamics simulations, large-scale language model (LLM) training, quantum chemistry, and generative AI models that generate new molecular structures for potential drugs. Tokyo-1 users can also take advantage of large language models for chemical, protein, DNA and RNA data formats with NVIDIA BioNeMo drug discovery software and services.
Mitsui’s AI-based drug discovery subsidiary Zeureca will operate Tokyo-1, which is expected to go live in the second half of this year. The initiative also includes workshops and technical training on accelerated computing and AI for drug discovery.
Japan’s pharmaceutical environment has long suffered from a drug lag, delaying the development and approval by therapists of drugs already in use elsewhere. The issue has received renewed attention during the race to develop a vaccine during the COVID-19 pandemic.
Japanese pharmaceutical companies see the introduction of AI as one of the solutions to this problem. In other words, AI is identified as a key tool to strengthen and accelerate the industry’s drug discovery pipeline. Training and developing AI models for drug discovery requires huge computing resources, such as the Tokyo-1 supercomputer. The first iteration of the Tokyo-1 supercomputer will feature 16 NVIDIA DGX H100 systems, each equipped with eight NVIDIA H100 Tensor Core GPUs.
The DGX H100 is based on the powerful NVIDIA Hopper GPU architecture and features a Transformer engine designed to accelerate the training of Transformer models, including generative AI models for biology and chemistry. Zeureca plans to add more nodes to the system as the project grows.
Customers can access the supercomputer’s dedicated servers, receive technical support from Zeureca and Nvidia, attend workshops from both companies, and request access to servers with more nodes for larger training sessions requiring more resources. computers than there are. Zeureca’s software solutions for molecular mechanics, docking, quantum chemistry and free energy perturbation calculations are also available for purchase.
Additionally, using NVIDIA Bionemo software on the Tokyo-1 supercomputer, state-of-the-art AI models can be tuned to billions of parameters for applications such as protein structure prediction, small molecule generation, and estimating the pose prediction.
Major Japanese pharmaceutical companies, such as Astellas Pharmaceuticals, Daiichi Sankyo and Ono Pharmaceuticals, are already planning to develop new drug development projects with Tokyo-1.
editor@itworld.co.kr
