High-Throughput Computing for Genomics and Bioinformatics Research

生物医学研究人员在全面研究细胞和分子生物学方面的能力越来越强，

Biomedical researchers have an increasing ability to comprehensively interrogate cell and molecular biology, for example with advanced imaging and next-generation sequencing. These techniques are producing datasets of increasing size and complexity and necessitate high-performance computing for biomedical researchers. Results from these studies are rapidly changing our understanding of normal development, our taxonomy of disease, and ultimately will enable precision medicine. Recognizing the importance of data science in precision medicine, the Institute for Precision Medicine (IPM) and the University of Pittsburgh invested in a high-throughput computing cluster (HTC), with infrastructure suited for memory-intensive and IO-intensive genomic and bioinformatic operations, and an education program dedicated to enabling computing for health science researchers. The HTC is located in the Center for Research Computing (CRC), a secure and centralized enterprise facility providing computing resources to the whole of the University. The CRC provides a scalable model for research computing with expertise in use and monitoring of advanced computing. In 2015, the HTC was installed with 16 nodes (256 cores). Usage of the HTC grew rapidly (2016 - 0.5M core-hours to 2019 – 3.1M core-hours), and despite the addition of computing nodes in 2017 (4 nodes), 2018 (4 nodes) and 2019 (16 nodes) the system often runs at capacity. Furthermore, the nodes have reached their useable lifespan. In this application, we are requesting funds to replace the HTC cluster and increase computing availability with a new cluster of 20 CPU nodes (960 cores) and 8 GPU cards, 2 CPU nodes as application servers, and 1.2 PB of BeeGFS storage, all linked by Infiniband HDR networking. The new HTC cluster is dedicated to NIH-funded research in biomedical science including medicine, pediatrics, cancer, immunology, and other areas. To assist health science researchers, the IPM, CRC and the health science library system (HSLS) support institution-wide licenses from numerous analysis and bioinformatics software suites, several of which are installed on the HTC server. Importantly, HSLS and CRC provide education outreach through workshops for command line tools, R for genomics and software suites. Researchers using software suites such as CLC Genomics, Open OnDemand, or command line can transparently migrate data and perform analysis on the multi-core HTC. IPM collaboration with the CRC ensures that the new HTC cluster will be operated in the most cost-effective and efficient manner and will be of high value to health sciences researcher performing NIH-supported research. This collaborative effort significantly reduces barriers for biomedical researchers needing to perform computational research.