AI Mount Sinai: High-performance computational and data ecosystem for biomedical discovery

Abstract: Affordable and available modern computational and data resources for AI-driven biomedical research projects are in short supply. AI Mount Sinai (AIMS) addresses this gap by providing state-of-the-art GPU capability and capacity for 42 research projects at 26 institutions. Research areas include: (1) AI for medical imaging and multi-modal clinical data, (2) structural and chemical biology, and (3) genetics and genomic sciences. Resulting research products will be disseminated through publications and national data repositories, including but not limited to: dbGAP, TOPMed, AMP-AD, CommonMind, PsychENCODE, BRAIN, Autism Sequencing Consortium and GEO, with the overarching goal of catalyzing translational science. AIMS consists of six Lenovo ThinkSystem SR780a V3 servers for a total of 48 NVIDIA B200 GPUs connected via NVLink and Infiniband NDR400. There are 9 terabytes of memory available on the Graphics Processing Units (GPUs), and an additional 12 terabytes available on the server nodes. Local high-speed NVME storage and DDR5 RAM enables caching of intermediate results by 672 compute cores on the compute nodes. Other features include a job scheduler, parallel file system and archival storage system. Mount Sinai is supplying networking infrastructure, short- and long-term storage and other essential supplies. Five years of operating costs, including expert personnel, facilities (power, cooling and space), hardware and software maintenance is also being funded by Mount Sinai to meet the acute scientific need for AIMS. Expert PhD and MD computational and data scientists' collaborate with biomedical researchers through extensive training and researcher engagement activities to maximize researcher productivity. The broad, long-term objectives of AIMS focus on improving our collective understanding of human disease in the following areas: Alzheimer's, autism, schizophrenia and related behavioral disorders, cardiovascular disease, kidney disease, diabetes type 2, drug addiction, depression and cancer. Without AIMS, there would be slower progress towards the diagnosis and treatment of these conditions affecting human health.