A Mitsui & Co. subsidiary aims to accelerate medication development by using powerful models built on data sets that can be safely shared thanks to the NVIDIA AI platform.
Since its founding 77 years ago, Mitsui & Co. has prospered by utilizing cutting-edge technologies such as generative AI and confidential computing to build ecosystems and businesses.
Digital revolution takes several forms at the Tokyo-based company with 16 divisions. In one case, it could be an autonomous trucking service, and in another, a geospatial analytics platform. Actually, Mitsui collaborates on cutting-edge quantum computing projects with a partner.
Xeureka, a new subsidiary, aims to expedite the ten-year, billion-dollar drug development process in the healthcare sector.
“It uses new digital technology like AI and confidential computing to create businesses,” said Katsuya Ito, project manager for digital transformation at Mitsui. * The majority of their work is done in collaboration with IT companies, such as NVIDIA and San Francisco-based security software company Fortanix.
In Search of Big Data
Even though Xeureka is just three years old, it has already completed a proof of concept that addresses one of the primary challenges in drug discovery: obtaining enough data.To accelerate drug development, powerful AI models that are built using datasets larger than those that are accessible to the majority of pharmaceutical companies are required. Up until recently, it was unthinkable for businesses to share data because it frequently contains private patient information and patented chemical compositions from pharmaceutical companies.
The practice of processing data in a safe section of a GPU or CPU that acts as a "black box" for a company's most important data is known as confidential computing.
The system, backed by a consortium of some of the world's largest companies, is used by banks, government agencies, and even advertising to ensure that their data is always kept private.
A Privacy Proof of Concept
To verify that private computing will allow its clients to safely share data, Xeureka created two fake companies, each with a thousand drug candidates. To predict the toxicity levels of the substances, an AI model was trained using the datasets of each company separately. The pooled data was then used to build a larger but similar AI model.Xeureka evaluated on NVIDIA H100 Tensor Core GPUs using security management software from Fortanix, one of the first companies to offer hidden computing.
Without compromising speed, the NVIDIA H100 GPUs provide a trusted execution environment with hardware-based engines that ensure and validate that private workloads are protected while being run on the GPU. The Fortanix program handles workflow management, encryption keys, and data interchange.
A 74% Increase in Accuracy
The outcomes were remarkable. The larger model made 65-74% more accurate predictions using the pooled datasets.Ito claims that the models created with data from a single company lacked the stability and bias found in the larger model.
“Confidential computing from NVIDIA and Fortanix basically alleviates privacy and security concerns while improving model accuracy, which will prove to be a win-win situation for the entire industry,” said Hiroki Makiguchi, CTO of Xeureka, in a news release from Fortanix.
An Ecosystem of AI Supercomputing
Xeureka is now exploring a variety of applications for their GPU-accelerated AI supercomputer, Tokyo-1, in drug discovery research in collaboration with the community behind it. Tokyo-1, which was revealed in February, aims to increase the productivity of pharmaceutical companies both domestically and internationally.Early initiatives could include collaborations to search ligand-base pairs, forecast protein structures, and expedite molecular dynamics simulations with trustworthy services.Tokyo-1 users can apply big language models for chemical, protein, DNA, and RNA data using NVIDIA BioNeMo, drug discovery microservices, and framework.
In addition to powering Japan's $100 billion pharmaceutical industry, which ranks third globally after the US and China, it is part of Mitsui's broader strategic expansion ambition to develop healthcare software and services.
Xeueka will use AI to quickly screen billions of drug candidates, predict how advantageous compounds will interact with proteins, and simulate complex chemical processes.
.jpg)
0 Comments