#Japan #Tokyo #NTT DOCOMO #GPU Technology #Mitsui Knowledge

Mitsui Knowledge and NTT Docomo Collaborate on GPU-Powered Genome Analysis

Mitsui Knowledge Corporation, based in Minato, Tokyo, and NTT Docomo Business, located in Chiyoda, Tokyo, have embarked on an impressive joint verification project utilizing the national wide-area distributed GPU environment known as the "GPU over APN Testbed". This initiative aims to enhance productivity in genome analysis within the bioinformatics sector.

In this project, Mitsui Knowledge has combined its advanced workload optimization techniques, including storage access, with high-speed search technologies for genome analysis data. This collaboration marks the first experimental use of the GPU over APN Testbed and involved executing a large-scale genome data analysis within an advanced dispersed GPU environment. Remarkably, the results of the trials demonstrated a time reduction in the analysis process of approximately 25% to 29%.

Background of Genome Data Utilization

Recent legislative frameworks, such as the Genome Medical Advancement Act and amendments to privacy protection laws, have fostered the utilization of genomic data across Japan. However, the rapid accumulation of this data presents challenges related to a shortage of human and computational resources in research environments. The escalating costs and constraints on high-performance computational resources, including GPUs, memory, and storage, further complicate these challenges.

Specifically, queries and analyses involving PB (petabyte) scale genomic data can take several days in traditional computing environments. The demand for accelerated verification cycles and enhanced productivity for researchers has thus become critical. In response to growing demands for GPUs driven by advancements in generative AI and data utilization, limitations such as capacity and power constraints in single data centers have surfaced.

To address these challenges, NTT Docomo Business has been validating distributed GPU technology that enables integrated use of geographically dispersed GPUs through the low-latency, high-capacity communication enabled by IOWN® APN. By July 2026, they plan to roll out a GPU over APN Testbed capable of processing computational tasks using GPUs distributed across eight locations nationwide.

Overview of Joint Verification

This joint demonstration project utilized the nationwide GPU over APN Testbed provided by NTT Docomo Business, paired with technologies developed by Mitsui Knowledge. The project executed genome analysis use cases distributed over five locations from Sapporo to Fukuoka, showcasing not just performance verification for high-speed searches, but actual applicability to research workload scenarios.

Project Objectives:

• - To validate the benefits of distributed execution by contrasting analysis workloads executed at a single location with those executed across five regions.

Use Cases:

1. Identifying specimens with mutations in specific chromosomal regions.
2. Aggregating the number of specimens with genetic traits linked to certain disease risks.
3. Identifying specimens with 220 known driver mutations linked to cancer onset.

Genome Analysis Data:

Utilizing publicly available genomic data samples covering 3,202 individuals from the International Genome Sample Resource (IGSR), which represents approximately 17 billion records and around 5TB of Variant Call Format genomic data.

Results of the Demonstration

The project successfully compared processing times between the single location analyses and the distributed processing across five sites. Findings indicated a significant analysis time reduction of 25% to 29%.

Future Prospects

Through their collaboration, both companies aim to accumulate insights towards the practical application of distributed GPU technology. They are focused on enhancing computational resource utilization in data-intensive research fields such as genome analysis. This initiative is set to bolster research activities in hospitals, universities, and research institutions nationwide while pushing further advancements in data utilization.

Furthermore, Mitsui Knowledge plans to continue research and development focused on the efficient use of increasing genome analysis data based on the insights gathered from workload optimization and high-speed search technologies. Concurrently, NTT Docomo Business will promote the creation of use cases in large-scale data analysis across various industries, leveraging the knowledge gained through distributed computing technologies for GPU as a Service (GPUaaS) offerings, thus aiming to establish a robust foundation for extensive AI processing capabilities.