Hybrid GPU/CPU Computing Resource to Support Connectomic and Genomics

University Of Maryland Baltimore, Baltimore MD

Investigators

Linked publications & trials

Abstract

CONNECTOMICS AND GENOMICS RESEARCH DESIGNATED HYBIRD GPU/CPU CLUSTER Project Summary/Abstract We request a high performance hybrid GPU/CPU computing cluster (HPC) to support the growing collaboration of connectomics and genomics studies at the University of Maryland School of Medicine (UMSOM). The equipment will be shared across the UMSOM. This inter-departmental resource will support a large portfolio of research projects among scientists at the Maryland Psychiatric Research Center (MPRC) in the Department of Psychiatry, the Department of Medicine Division of Endocrinology, Diabetes & Nutrition (DOM-EDN), and the Institute for Genomic Studies (IGS). The research programs within these centers are collaborative, complementary, and focused on analyses of high dimensional phenotype and genotype data. The combined research portfolio includes thirty one eligible projects (U01/54, P30/50/54 and R01 grants) separated into seventeen major and fourteen minor users. We highlight the recently awarded Amish Connectome Project (ACP) in Mental Illness that was funded through a Human Connectome in Disorders NIH initiative and Adolescent Brain Cognitive Development (ABCD) projects as examples of collaborative projects undertaken by our multidisciplinary science team. The ACP and ABCD projects high-dimensional connectomics and genomic data will be co-analyzed to identify genomic-connectomic-disorder pathways. Many similar projects flourish and cross-pollinate through a dedicated and shared computational resource support in our campus. The recent expansion of our research portfolio has prompted the need to upgrade of existing capabilities. Our team is uniquely position to operate such a resource because of its NIH-funded scientific software development methods for high-performance computing. The hybrid GPU/CPU analyses methods for imaging and genetic analysis software capable of performing massively parallel genomic analyses in imaging phenotypes to be tested using this resource will benefit researchers in other institutions that rely on our tools. Given the synergy and collaborative nature of our research we seek an HPC that can meet our collective needs rather than upgrading the current computing infrastructure for each program individually.