Science Success Story
XSEDE announces 2020-2021 Campus Champions Fellows
Six researchers selected to partner with XSEDE experts for year-long collaborations
By Boswell Hutson, National Center for Supercomputing Applications (NCSA)
Six members of the 700+ member Campus Champion community have been selected as Campus Champions Fellows for the 2020-2021 academic year and will have the opportunity to work side-by-side with XSEDE staff and research teams to solve real-world science and engineering projects.
The six Fellows selected for this year will work on projects spanning bioinformatics to image analysis of historic films under the overarching goal of increasing cyberinfrastructure expertise on campuses by including Campus Champions as partners in XSEDE's projects. While all six Fellows chose to work with XSEDE's Extended Collaborative Support Services (ECSS) staff, the program had been expanded over the years to span all of XSEDE and has recently included Workforce Development, XSEDE Cyberinfrastructure Integration and Cybersecurity. The Fellows and their projects are as follows:
Fellow: Suxia Cui, Prairie View A&M University
Mentor: Alan Craig, University of Illinois and Shodor
ECSS PI: Greg Wilsbacher, University of South Carolina
Project Title: Image analysis for digital surrogates of historical motion picture film
Project Description: For almost a century, celluloid-based imagery was the dominant medium for recording the history of the world, creating a global library of still and moving images the full epistemological weight of which has yet to be felt. Even though still image digitization has reached a mature state, the archival digitization of motion picture film is still in a developmental phase because motion picture film is derived from a complex system and results in data intensive files. But given the rise of deep fake technology it is essential that mature systems are developed soon. Moving Image Research Collections (MIRC) at the University of South Carolina recently entered into a partnership with the United States Marine Corps History Division to preserve, digitize and make accessible the legacy 16mm and 35mm film collection housed at Marine Corps University, Quantico. The collection is very large, containing over 18,000 cans of film (a typical can contains 7 to 8 minutes of footage). MIRC has been scanning films at 2K (2048 x 1531 pixels) and currently has over 2,000 digitized films, typically scanning 60 to 75 cans per week. Digitizing the films is only one component of the project. This collection has a high research value for historians of many types and is of sufficient size to create a data set able to train image analysis algorithms. MIRC seeks to identify new methods for deploying these digital film assets as trusted historical resources (in contrast to the chaos of user-contributed online video). To accomplish this, MIRC is partnering with the university's Computer Vision Lab (led by Dr. Song Wang) and its Research Computing unit to develop and deploy three initial projects: Machine Learning (ML) algorithms for identifying and tracking textual information in historical imagery; ML algorithms for facial recognition in historical imagery; and a new method for certifying the chain of historical provenance from a celluloid film to a master digital surrogate copy, and then to all subsequent copies derived from that master. MIRC seeks support to build a virtual home that not only hosts the online collections for the public, but also allows researchers and developers to collaborate with others and experiment the mechanisms for above mentioned image/video analysis projects.
Fellow: Kurtis Showmaker, University of Mississippi Medical Center
Mentor: Choonhan Youn from the San Diego Supercomputer Center (SDSC)
ECSS PI: Guoshuai Cai, University of South Carolina
Project Title: A Comprehensive Annotator and Web Viewer for scRNA-seq Data
Project Description: Individual cells are the building blocks of tissues, organs, and organisms. Each tissue contains cells of many types, and cells of each type can switch among biological states. To understand how this complex system work, it will be important to learn the functional capacities and responses of each cell type and each single cell. A major determinant of each cell's function is its transcriptional program, which can be studied by single cell transcriptomics analysis. Recently a powerful technique, single-cell RNA-sequencing (scRNA-seq), has been developed and enables high-resolution studies of gene expression patterns at the single-cell level. However, its high data volume and complexity bringing in many new computational challenges, including the effective ways of data visualization as well as the comprehensive biological annotation. To fulfill this urgent demand, we are developing a web application for RNA-seq data annotation and visualization, which need the XSEDE computing resource as requested. We believed that this application will have significant impact in single cell research in biomedical area. We expected to publish this tool with a series of useful features in the journal of Bioinformatics or Nucleic Acid Research for public use.
Fellow: Nitin Sukhija, Slippery Rock University of Pennsylvania
Mentor: Paul Rodriguez, San Diego Supercomputer Center (SDSC)
ECSS PI: Toni Whited, University of Michigan
Project Title: Estimating Dynamic Models of the Firm
Project Description: Research on estimating dynamic models of the firm has become a fixture of empirical financial economics. However, development has been hampered by the computational burden involved in solving a complex dynamic program, tens of thousands of times. We are requesting a renewal of our allocation to cover our agenda in the study of dynamic models, which would otherwise be impractical to pursue. We continue to investigate several new models of firm behavior. The topics include monetary policy transmission and international finance and growth. We have developed a set of scalable programs on XSEDE computers that allow this agenda to make full use of the resources available through an XSEDE allocation.
Fellow: Zhiyong Zhang, Stanford University
Mentor: Hang Liu from the Texas Advanced Computing Center (TACC)
Fellows designed project: Optimal Utilization of XSEDE Computing Resources for the NWChem Computational Chemistry Software Package
Project Description: NWChem is one of the most popular molecular chemistry simulation packages and is specifically designed to make optimal use of massively parallel supercomputers. The software has a large user base and is available on most XSEDE resources and many campus computing facilities. It contains all the essential functionalities of modern electronic structure theory and new capabilities continue to be added. Moreover, it is a primary candidate for porting to the next generation Exascale computers. Campus Champion Fellow Zhiyong Zhang (Stanford) will work with ECSS scientific computing expert Hang Liu (TACC) on a Fellows-designed project to improve the scalability and performance of NWChem on a variety of architectures to take advantage of the latest generations of processors, accelerators, interconnects and NVMe devices. The emphasis will be on XSEDE resources, such as Stampede2, Bridges2 and Comet. Specifically, they will carry out roofline model analyses, which incorporate performance, memory bandwidth, and memory locality metrics, to understand the trade-off between data-movement and computation. The goal is to identify the bottlenecks (cache, memory, local and remote I/O, internode communications) that limit performance and reduce their impact.
Fellow: Sinclair Im, Yale University
Mentor: Vinit Sharma, National Institute for Computational Sciences (NICS)
ECSS PI: Pratibha Dev, Howard University
Project Title: A density functional theory study: quantum materials
Project Description: Understanding how neural circuits mediate animal behaviors is a fundamental problem in neuroscience, one that requires an inventory of the cell types comprising these biological circuits and experimental access to specific cell types to elucidate their roles in neural circuit function. With the advent of high-throughput single cell transcriptome profiling, the past few years have witnessed an explosion of new information on the complexity of cell types in the nervous system based on genes expressed by individual cells [reviewed in ref. 1]. In addition to providing a principled basis with which to create a taxonomy of cell types in the brain, knowledge of the genes expressed by specific neurons also provides information to target genetically encoded reporters and actuators to these newly discovered cell types . An avalanche of data is expected from research consortia - including the BRAIN Initiative Cell Census Network - whose goal is to create a comprehensive census of cell types in the mouse brain using single-cell RNA-sequencing (scRNA-seq) approaches. New statistical methods will be needed to analyze these large datasets - containing upwards of millions of cells - to classify cells based on their transcriptomes and identify biomarker genes that can be used to identify and interrogate them experimentally. We will develop new statistical approaches for data normalization, clustering and biomarker identification that can be scaled for analyzing large scRNA-seq datasets. As a related case, we will be working with single-cell RNA-sequencing data from the olfactory epithelium. We have previously used scRNA-seq to understand how olfactory stem cells contribute to the remarkable regenerative capacity of this neurogenic tissue [3, 4]. We are currently revisiting these experiments at higher cellular resolution, to better distinguish different cell types in the regenerating tissue. These will provide an additional dataset for the development of normalization and clustering methods. References: 1. Poulin JF, Tasic B, Hjerling-Leffler J, Trimarchi JM, Awatramani R. Disentangling neural cell diversity using single-cell transcriptomics. Nat Neurosci. 2016;19(9):1131-41. 2. Huang ZJ. Toward a genetic dissection of cortical circuits in the mouse. Neuron. 2014;83(6):1284-302. PMCID: PMC4169123. 3. Fletcher RB, Das D, Gadye L, Street KN, Baudhuin A, Wagner A, Cole MB, Flores Q, Choi YG, Yosef N, Purdom E, Dudoit S, Risso D, Ngai J. Deconstructing Olfactory Stem Cell Trajectories at Single Cell Resolution. Cell Stem Cell 2017;20(6):817-830.e8. 4. Gadye L, Das D, Sanchez MA, Street K, Risso D, Baudhuin A, Cole MB, Wagner A, Choi YG, Purdom E, Dudoit S, Yosef N, Ngai J, Fletcher RB. Injury Activates Transient Olfactory Stem Cell States With Diverse Lineage Capacities. Cell Stem Cell 2017;21(6): 775-790.e9.
Fellow: Brady Butler, University of Maine
Mentor: Manu Shantharam, San Diego Supercomputer Center (SDSC)
ECSS PI: Fazle Hussain, Texas Tech University
Project Title: Numerical study of supersonic turbulent boundary layer drag control and vortex reconnection cascade at high Reynolds numbers
Project Description: This proposal is a request for computer time on TACC STAMPEDE2 (SKX nodes) along with archival storage on TACC RANCH, to be used in our ongoing studies of active turbulent skin-friction drag reduction (DR) and vortex reconnection cascade. For drag control, we would perform direct numerical simulation of drag control with spanwise wall oscillation for supersonic turbulent channel flows at bulk Reynolds number Reb = 17000 and bulk Mach number Mb = 1.5 and 3. The new results, in combination with our previous study at lower Reynolds and Mach numbers, would enable us to obtain a clear Reynolds and Mach number scaling, which is of great importance to evaluate the capability of drag control method at high practical situation. For the vortex reconnection studies, the main goal is to improve the parallel efficiency of our code so that we can be ready to perform simulation at high Reynolds numbers, which is essential for understanding the vortex reconnection cascade mechanism and its relationship to turbulence cascade.
Accepted Fellows, with the support of their home institution, make a 400-hour time commitment and are paid a stipend to allow them to focus time and attention on these collaborations. The program also includes funding for two visits, each ranging from one to two weeks, to an ECSS, PI or conference site to enhance the collaboration.
For more information on the XSEDE Campus Champions Fellows program, including all past cohorts, visit: https://www.xsede.org/ccfellows.