Computational Pathology High Performance Computing Cluster

Project Summary/Abstract Genomics, digital imaging, and spatial biology technologies generate a large amount of data that needs to be processed and analyzed with a high-performance computing (HPC) cluster. Interpretation of these high dimensional data could be significantly enhanced with artificial intelligence which also requires extensive data storage, high-speed network for data transfer, and substantial CPU/GPU capacity for model training and inference. The existing HPC cluster managed by Mass General Brigham (MGB) Enterprise Research Information Systems (ERIS) is outdated with old technology that is further exacerbated by an oversubscription of ERIS users competing for limited resources in the cluster. An alternative research cluster in the Massachusetts Host-Microbiome Center (MHMC) also faces similar issues with storage, networking, and limited GPU capacity. Recognizing these challenges, we propose to build and support the MGB Pathology Research HPC cluster with ten CPU nodes, seven petabytes of resilient TrueNAS storage, five H100 GPU nodes, and vast NVMe storage technology all connected with a fast 200GbE ethernet storage networking and a modern 400GbE Infiniband fabric for GPU networking. We will use a software layer with Rancher, Harvester, Longhorn, and Kubernetes to further ensure high availability, load balancing, and ease of application orchestration. This proposal has garnered the full support of MGB Digital and MGB Pathology leadership who have committed in-kind salary support for an HPC Specialist, an Infrastructure Team Lead, and an Infrastructure Engineer to help design, build, implement, and support the MGB Computational Pathology HPC Cluster. Additionally, in-kind departmental funds under the discretion of the PI and Co-I of this S10 High-End Instrumentation Grant (HEI) will further ensure success for this project. This proposed platform will accelerate numerous Mass General Brigham NIH-funded projects in genomics, spatial imaging, digital pathology, and AI, thereby driving innovation and discovery in these fields as exemplified by the 21 projects proposed by 6 major and 4 minor users. The PI, Co-I, and Instrument Advisory Committee will ensure that this system will be available to the broader MGB Pathology community and collaborators in the Greater Boston region. This will enable high-performance computing for NIH-funded projects, thereby fostering a collaborative and innovative research environment.