Supercomputing and Life Sciences Symposium 2019


April 17, 2019

Symposium

9:00 am to 5:00 pm
Willa Cather Dining Center
530 North 17th Steet
Lincoln, NE

Reception and Poster Session

5:30 pm - 7:30 pm
Morrill Hall
645 North 14th Street
Lincoln, NE

Register Now

The Holland Computing Center and the Quantitative Life Sciences Initiative have partnered to host the 2019 Supercomputing and Life Sciences (SLS) Symposium, bringing you speakers on computational sciences from across the nation and within the NU system.

Presentations will be evaluated and awarded prizes based on attendee votes. For more information, check out the "Presenter Information" tab below.

Symposium Agenda

9:00 - 9:30 Check In and Coffee
9:30 - 9:50 Opening Remarks
9:50 - 10:30 David Warren, Assistant Professor
Department of Neurological Sciences - University of Nebraska Medical Center
Slides
10:30 - 10:50 Break
10:50 - 11:30 Jonathan Bentz, Solutions Architect Manager
NVIDIA
Slides
11:30 - 12:00 State of HCC
David Swanson, Director
Holland Computing Center - University of Nebraska
Slides
12:00 - 1:30 Lunch / VR Expo / Oversight Committee Meeting
1:30 - 2:10 Andrew Benson, Director
Nebraska Food for Health Center
2:10 - 2:30 Determining putative bioactive peptides: how we used the Open Science Grid
Jean-Jack Riethoven, Director
Bioinformatics Core Research Facility - University of Nebraska
2:30 - 2:50 Opening new research avenues by creating links between disparate data repositories
Lucas Busta, Post Doctoral Research Fellow
Biochemistry - University of Nebraska
AbstractSlides
2:50 - 3:10 Break
3:10 - 3:50 Computing at Los Alamos National Laboratory: HPC in Transition
Cory Lueninghoener, HPC System Integration Team Lead
High Performance Computing - Los Alamos National Laboratory
AbstractSlides
3:50 - 4:30 Making mountains out of molehills: Challenges for implementation of cross disciplinary research in the big data era
Daniel Andresen, Director
Institute for Computational Research
AbstractSlides

Evening Reception Agenda

5:30 - 6:45 Reception and Poster Voting
6:45 - 7:25 Planetarium Show / Poster Vote Tally
7:25 - 7:30 Closing Remarks and Prize Announcements

Attendee Information

Click Here to Register

Registration ends April 10th!

SLS Symposium

Map and Directions:

The SLS Symposium will be held at the Willa Cather Dining Center on the University of Nebraska-Lincoln City campus.

Willa Cather Dining Center
530 N 17th St
Lincoln, NE 68588

Parking Information:

Parking is subject to UNL's Parking Regulations. Reciprocal permits are available for NU assocates with permits on their home campuses.

Public parking is availble in the form of metered parking along 19th Street. Hourly garage parking is available at the 19th and Vine garage, and UNL Permit lots are also available and can be found here.

Reception

Map and Directions:

The Reception will be held at the University of Nebraska State Museum in Morrill Hall on the University of Nebraska-Lincoln City campus.

Nebraska State Museum
645 North 14th Street
Lincoln, NE 68588

Parking Information:

Parking is subject to UNL's Parking Regulations. Reciprocal permits are available for NU assocates with permits on their home campuses.

Public parking is availble in the form of metered parking along 14th Steet and along T street. Hourly garage parking is available on Stadium Drive and on 14th and Avery. UNL Permit lots are also available and can be found here.

Presenter Information

Click Here to Register as a Presenter

The Holland Computing Center (HCC) and the Quantitative Life Sciences Initiative (QLSI) are hosting the Supercomputing and Life Sciences Symposium on April 17, 2019. We invite poster presentations covering research relevant to the QLSI mission and/or benefiting from advance computing (e.g. use of HCC, the Open Science Grid (OSG) or XSEDE resources). While this session is open to anyone with relevant research, current graduate students are strongly encouraged to participate.

Prizes

Top presenters will be chosen based on the quality of research and their use of HCC services and/or application of Big Data in Life Sciences.

Prizes will be presented to Graduate Student participants, based on attendee votes.

Featured Prize List:

  • Dell Latitude 7490 Laptop: Featuring a i7-8650u Processor, 16GB DDR4 Memory and 512GB NVMe Storage
  • New* 40" UHD SmartTV
  • Lenovo Mirage AR Headset featuring the Star Wars: Jedi Challenges Smartphone Augmented Reality Experience

No need to create a new poster, presenters welcome to use existing posters.

Interested individuals should register and submit abstracts by April 7th to be considered.

* The television will be used for HCC's VR Expo over lunch, and will be returned to the original packaging for the winner

Presentation Guidelines

  • Prizes will go preferentially to research that utilizes advanced computing resources such as HCC, XSEDE, OSG and/or features the application of Big Data in Life Science.
  • Any topic related to advanced computing or big data, however, is welcome. A main goal is to spur and increase discussion and potential collaboration!
  • Posters should be no more than 48" x 48" in size. For posters larger than this, please contact us at hcc-support@unl.edu to make arrangements.

Speaker Abstracts

Opening new research avenues by creating links between disparate data repositories

Lucas Busta
Post-Doctoral Researcher - Univeristy of Nebraksa-Lincoln

The past decades have seen massive changes in the landscape of scientific research. The internet now provides scientists in diverse fields with the ability to create large databases that organize and make available astounding amounts of information. In parallel, the internet has fostered many areas of modern research that take place at the intersection of multiple classical research fields by enabling teams of researchers with distinct skillsets to communicate and combine their datasets. While it is relatively commonplace for researchers to use the internet to create links between themselves and thereby facilitate these collaborative projects, the creation and exploitation of such links between scientific databases seems to occur much less frequently. Here, I describe the challenges, payoffs, and insights from a research project that combines a large chemical database (CAS SciFinder), taxonomic databases (Tree Of Life and The Plant List), and a large genomic database (JGI Phytozome).

Computing at Los Alamos National Laboratory: HPC in Transition

Cory Lueninghoener
HPC System Integration Team Lead - Los Alamos National Laboratory

Scientific computing at Los Alamos National Laboratory has a history stretching back to the 1940s, and throughout its history the laboratory has been home to some of the fastest and most innovative computing systems in the world. But with such a long computing legacy comes a long software legacy, resulting in growing pains as we embrace modern technologies for our next generation system software stack. This talk will take a brief look at the HPC systems at LANL and the science they support, and then dive more deeply into ways that we are changing our system software design to support higher flexibility, efficiency, and scalability using technologies like containers, exascale workflows, and distributed system automation tools.

Making mountains out of molehills: Challenges for implementation of cross disciplinary research in the big data era.

Daniel Andresen
Director - Institute of Computational Research

In this talk we present a Researcher’s Hierarchy of Needs (loosely based on Maslow’s Hierarchy of Needs) in the context of interdisciplinary research in a “big data” era. We discuss multiple tensions and difficulties that researchers face in todays environment, some current efforts and suggested policy changes to address these shortcomings, and present our vision of a future interdisciplinary ecosystem.

Poster Abstracts

Dynamic Fracture at an Interface: a Peridynamic Analysis

Javad Mehrmashhadi, Longzhen Wang, Quang Le, Florin Bobaru

Recent impact experiments showed the influence a strong or weak interface in a bi-layered PMMA material has on dynamic fracture mechanisms. We show that a linear elastic with brittle damage peridynamic model, which works very well for glass, leads to crack propagation speeds significantly faster than those measured experimentally in the PMMA system. We propose an explanation for this behavior: localized heating in the region near the crack tip (due to high strain rates) softens the material sufficiently to make a difference. We introduce this effect in our peridynamic model, via a bi-linear bond force-strain relationship, and the computed crack speed matches the experiments. We perform a detailed peridynamic analysis concerning the effect a strong or weak- interface in the bi-layered PMMA system has on how long the crack runs along the interface before penetrating in the second layer. The computed crack paths follow very closely those found experimentally. The proposed model seems to explain the observed behavior.

Coat proteins in Potyviruses are Variable.

Katherine LaTourrette, Deepti Nigam, and Hernan Garcia-Ruiz

The genus Potyvirus consists of 167 species that can infect a wide variety of wild and domesticated plants. Their polyprotein consists of several different genes including the coat protein (CP). Previous studies have found that the Asp-Ala-Gly (DAG) motif located on the CP is required for aphid transmission. Using various bioinformatic approaches, this study will determine the location of variable regions, positive selection sites, and DAG motifs in the potyvirus CP. Structural alignment results show that, for potyviruses, the CP forms a variable N-terminal loop, a core, and a C-terminal loop. After normalizing for the length of each region, the variable N and C terminal loops harbor more variation than the core. The DAG motifs map to the N terminal loop and are not conserved. There is variation in both location and sequence. Emphasizing this result, the Bean Yellow Mosaic Virus CP did not contain a DAG motif. These results imply disorder is a general feature of the potyviral CP. This suggests a correlation between variation in CP, aphid-transmission, and virus-vector specificity. Potentially, genetic flexibility in potyviral genomes is necessary to maintain functionality in genetically diverse host and vectors.

Mechanism and Kinetics of Transition Metal Dissolution from Spinel LiNi0.5Mn1.5O4 Cathode Material

Nadia N. Intan and Vitaly Alexandrov

Irreversible transition metal (TM) dissolution of the cathode material in lithium ion batteries (LIBs) represents a serious challenge for the development of stable high energy density LIBs. One of the most extensively studied LiNi0.5Mn1.5O4 (LNMO) spinel structured cathode material also suffers from electro-active material (Mn and Ni) dissolution. The dissolution phenomenon of TM from cathode surface of LIB contributes to the capacity loss of LIBs and formation of solid electrolyte interface (SEI) over extended cycle period. We combine static and ab initio molecular dynamics (AIMD) free energy calculations within the density functional theory (DFT) to explore the interactions between electrolyte species and LNMO spinel surfaces. We specifically employ enhanced free-energy sampling techniques to simulate TM dissolution with explicit treatment of organic electrolyte to explore the dissolution mechanism and kinetics. By considering the effect of various species that maybe present at the electrode-electrolyte interface on dissolution barriers we obtain key insights into the dissolution process. Also, we establish a correlation between the adsorption strength and dissolution barrier that can be used to predict degradation behavior across LIB cathode/electrolyte interfaces.

On Peridynamic Modeling of Crack Nucleation

Sina Niazi, Ziguang Chen, and Florin Bobaru

Quasi-brittle crack propagation under static or dynamic loading has been successfully analyzed using peridynamics. However, existing peridynamic models based on a single fracture parameter, associated with critical fracture energy, produce different strengths for different nonlocal region (horizon) sizes in modeling crack nucleation under quasi-static conditions from sites other than pre-cracks. To introduce an independent parameter from the one linked to the critical fracture energy into the peridynamic bonds behavior, one can use a bi-linear bond force-strain relationship and relate the extra parameter to the ultimate strength of the material. Recent works focused on aspects of crack nucleation from a material stability point of view and crack propagation based on such bi-linear behavior of peridynamic bonds. In our study, we analyze the bi-linear model to study crack nucleation with the goal of connecting the parameters to measurable physical quantities, like nucleation of cracks in indentation of brittle samples. We perform convergence studies in terms of the nonlocal region going to zero and confirm results published in the experimental literature.

Algorithms for Updating Dynamic Network

Sriram Srinivasan

The growth of social media increased the interest in analyzing network algorithms. The networks are highly unstructured and exhibit poor locality, which has been a challenge for developing scalable parallel algorithms. The state-of-the-art network algorithms such as Prim's algorithm for Minimum Spanning Tree, Dijkstra's algorithm for Single Source Shortest Path and iSpan algorithm for detecting strongly connected components are designed and optimized for static networks. The networks which change with time i.e. the dynamic networks such as social networks, the abovementioned approaches can only be utilized if they are recomputed from scratch each time. Performing a re-computation from scratch for a significant amount of changes is not only computationally expensive, however, increases the memory footprint and the execution time. In the case of dynamic networks, developing scalable parallel algorithms is very challenging and there has been a very limited amount of research work that has been performed when compared to developing parallel scalable algorithms for static networks. To address the above challenges, this research proposes a new high performance, scalable, portable, open source software package and an efficient network data structure to update the dynamic networks on the fly. This approach is different from the naive approach which is the recomputation from scratch and is scalable for random, small-world, scale-free, real-world and synthetic networks. The software package currently is implemented on a shared memory system and updates network properties such as Connected Components (CC), Minimum Spanning Tree (MST), Single Source Shortest Path (SSSP), and Strongly Connected Components (SCC). The key attributes of the software are faster insertions and deletions. Additionally, the software takes less time and memory for updating the networks when compared to the state of the art Galois. The shared memory implementation processes over 50 million updates on a real-world network under 30 seconds.

Fate of Antimicrobial Resistance in the Environment: From Beef Cattle Production to Manure Storage and Land Application

Ece Bulut, Darshan Baral, Xu Li, Galen Erickson, Amy Schmidt, John W. Schmidt and Bing Wang

Knowledge gaps exist regarding the transfer of antibiotic resistant bacteria from livestock to humans via environmental pathways, which hinders a systems assessment of the impact of antibiotic uses during food-producing animal husbandry on public health. Therefore, the purpose of this study was to evaluate the survival and transfer of antibiotic resistant bacteria and genes in the environment in a continuum of beef cattle primary production, cattle manure storage and land application. In this study, different in-feed antibiotic treatments (Control, Tylan and Chlortetracycline) were introduced to beef cattle on feedlot (32 animals/treatment). Samples of rectal feces, hides and pen surface on feedlot (5 months); manure during stockpiling (3 months; 12 samples/pile/sampling); and amended soil at land application sites (3 months; 16 samples/site/sampling) were collected. Changes in prevalence and concentration of generic and macrolide- and tetracycline-resistant E.coli, Salmonella and Enterococcus were determined over the project lifespan. Organisms and antibiotic resistance genes were characterized and quantified following shotgun metagenomics sequencing using Illumina HiSeq platform. Our results indicated that there was no statistically significant difference in antibiotic resistant bacteria load and gene abundance was detected across antibiotic treatments throughout the study from beef cattle production to manure application. Taxonomic composition of samples were similar between treatments. During the 3-month period of manure storage as stockpiles, the concentration of generic E.coli and Enterococcus dropped from ~5 log10CFU/g to a maximum of 2-3 log10CFU/g. Manure storage as static piles significantly reduced antibiotic resistant bacteria and genes in three months. Our results indicate antibiotic use during beef cattle production might not be associated with an extra risk of contamination of antibiotic resistant bacteria and genes in animal wastes and following manure and amended soil. Stockpiling with sufficient time can effectively eliminate resistant bacteria and genes in manure, indicating sufficient manure intervention could effectively limit the release to the environment through land application and subsequently reduce the direct and indirect exposure to environment-mediated antimicrobial resistance in humans.

Bioinformatics Analysis To Find Significant Functional Genes In Antibiotic Responses In Staphylococcus Species

Kimia Ameri, Kathryn Cooper, Austin Nuxoll

Persister cells are dormant variants of regular cells which are responsible for making bacteria resistant to new drugs and multidrug tolerance. In this study, we provide an automated pipeline to distinguish significant genes between two types of isolates with different antibiotic responses. for this purpose, isolates need to be labeled as high or low persister type. This fully automated analysis starts with raw fasta file, and will give the most significant genes based on their response to antibiotics as an output. All the process will be run in crane cluster of HCC with its bioinformatics tools.

BADPEP: A computationally generated database of dietary peptides with putative antimicrobial and protein-protein interaction

Qidong Jia, Jean-Jack M. Riethoven, Jennifer Clarke, Rohita Sinha

Background: The gut microbiome has emerged as a pseudo organ due to its enormous biochemical potential and taxonomic diversity. Compositional abnormalities (dysbiosis) in this microbiome have been associated with multiple metabolic disorders and diseases. Thus, a natural approach to treat metabolic disorders is the systematic manipulation of gut microbial composition; dietary molecules, the predominant factors affecting the gut microbiota, are an obvious route to achieve this. Dietary peptides are one such class of dietary molecules with antihypertensive, antioxidant and anti-inflammatory activities. Result: Here we present the initial version of BADPEP (http://badpep.unl.edu/), a database of dietary peptides computationally screened for their antimicrobial activities and ability to competitively inhibit protein-protein interactions (PPIs). Inhibition of vital protein-protein interactions is a new drug development paradigm as few protein-protein interface fragments, also known as hot-segments, contribute most of the overall binding energy and can be used in the development of novel PPI inhibitors. Method: The computational pipeline behind the BADPEP database mimics the human gastrointestinal digestion of dietary proteins by simulating the activities of the enzymes Pepsin, Trypsin and Chymotrypsin. The resulting set of in silico digested peptide fragments were compared with a known set of antimicrobial peptide (AMP) sequences and hot-segments of known protein-protein complexes. Dietary peptides with significant similarities with known AMPs and PPI interface-fragments can be easily queried through the BADPEP database. The Holland Computing Center's clusters S\andhill, Ttusker, and Ccrane, together with the Open Science Grid, weare utilized in excess of 20.5 million CPU hours to support the computational aspects of this work. The underlying MySQL database itself is located on HCC’s cloud instance Anvil. Conclusion: In the current framework, users can query the BADPEP database by submitting the Uniprot IDs of known protein sequences or they can query for all the bioactive-peptides present in a plant or animal species. We hope that BADPEP will assist in elucidating the mechanism of action of bioactive-peptides and contribute to the design of novel experiments to evaluate the bioactivities of dietary proteins.

Computational approaches to predict neoantigens for improving cancer immunotherapy

Vi Dam, Dario Ghersi

The toxicity of conventional cancer treatments is not limited to cancer cells but it extends to healthy cells. Therefore, it is critically important to target cancer cells while minimizing damage to healthy tissues. Neoantigens are very promising targets in cancer immunotherapy because they are specific to cancer cells only. Targeting these tumor-specific neoantigens is ideal as the risk of adverse side effects and healthy cells death is minimized. However, due to the complexity of our immune system as well as the extreme diversity of neoantigens in cancer types and patients, only a fraction of patients is responsive to immunotherapy. Elucidating the mechanisms of tumor cell elimination by the immune system and identifying patient-specific neoantigens is therefore imperative for improving therapeutic outcomes. In this research, I investigate the use of computational methods to identify neoantigens in colorectal cancer and to predict the binding affinity of neoantigens to cell surface proteins. Further, this project aims to develop a computational framework for validating these predicted neoantigens, which is critical given a large number of false positive neoantigen candidates.

Corn and Soybean Root Microbial Communities Respond to Crop Rotation and Nitrogen Fertilization

Michael A. Meier, Qidong Jia

Crop rotation is a an agricultural practice that leads to higher net yield, increased disease tolerance, decreased weed densities and reduced need for nitrogen (N) fertilization. It is thought that root-colonizing soil microbes are partially responsible for these effects but it is poorly understood how different crop rotation systems and N fertilization regimes affect root microbial communities. UNL has been conducting long-term field experiments in collaboration with USDA-ARS since the late 1970’s. In some experimental plots consistent crop rotations and nitrogen fertilizer regimes have been in place for close to four decades. In this two-year study we analyzed root and soil DNA samples taken from long-term continuous corn (Zea mays L.), continuous soybean (Glycine max L.), and corn/soybean rotation systems under low and high N regimes. Microbial diversity was assessed through amplification and Illumina sequencing of the 16S rDNA V4 region. Sequence data was analyzed in R (DADA2/PhyloSeq/DESeq2) using the computational resources at the UNL Holland Computing Center.

Interpersonal variation and susceptibility to antibiotic-resistant pathogens: Complimentary in vitro and in vivo models.

Armando Lerma, Kylie Farrell, Thomas Auchtung, Robert Britton, Anthony Haag, and Jennifer Auchtung

Antibiotic-mediated microbiome disruption and consequential reduction in levels of beneficial microbial metabolites could have potential detrimental implications in gut health. These include loss of resistance to gastrointestinal (GI) pathogens colonization, emergence of multi-drug resistant (MDR) organisms and driving emerging pathogens into new niches as it has been found for Clostridioides difficile infection (CDI). It is also known that interpersonal variation can influence susceptibility to antibiotic-mediated disruption. Developing pre-clinical models that allow us to evaluate the impacts of antibiotics in multiple configurations of the human gut microbiome could lead to discovering new therapies to restore its functions and prevent or reverse disease states. Two complimentary models of human fecal microbial communities, HMbmice and human fecal minibioreactor arrays (MBRAs), were rigorously tested with diverse fecal samples by being treated with different classes of antibiotics that are either known to lead to widespread disruption (Clindamycin and Cefaclor) or have limited impact on the microbiome (Fidaxomicin). 16S rRNA data demonstrated that both models have the potential to model complex patterns of resistance and susceptibility to disruption by antibiotics. The data also revealed a variety of responses to antibiotic consumption in diverse microbiomes showing the most robust microbiome composition disruption caused by clindamycin. Liquid chromatography–mass spectrometry detected metabolite level changes in both models with clindamycin having the biggest impact on bile acid levels. Finally, whole genome sequence analysis data presented high levels of antibiotic resistant genes in MBRAs but generally lower levels in HMbmice. The two models validated in this study used in combination could facilitate development of new therapies by providing physiologically relevant models of human GI microbial communities for pre-clinical testing.

Computer-aided discovery of fundamental deformation mechanisms of structural materials

Jian Wang

Understanding and predicting deformation behaviors of structural materials are the fundamental needs for designing and processing materials, and advancing the application of structural materials. Experiments at different scales offer a plenty of phenomena as the material is subjected to mechanical loading and provide insights into understanding deformation mechanisms and behaviors of materials. However, some deformation events and underlying mechanisms can not be captured using current experimental techniques, such as migration of interfaces, dislocation nucleation, and nucleation of deformation twins and phase transformation. We employed computation experiments with the support of superpower computers to explore deformation mechanisms of structural materials, and develop macro-scale computer modeling tools for predicting mechanical response of structures. In this talk, we report our work with focus on interfaces engineering.

Molecular Communication in Cell Metabolism Communication & Information-centric Computational Tool in KBase

Zahmeeth Sakkaff, Nidhi Gupta, Massimiliano Pierobon and Christopher Henry

In natural environments, bacteria interact with each other and generally exist as microbial communities. These communities play a central role not only in the cycling of nutrients like carbon and nitrogen but also in the degradation of complex polyaromatic compounds. With the evolution of next-generation sequencing technology, more and more microbial community sequencing data are becoming available from various environmental niches leading to the identification of the different microbial species present in the community. However, to understand the functional contribution of each individual species in the community and the overall functional differences between communities, community analysis via mathematical modeling and computational analysis is required. In the present study, we utilized information-centric computational approaches that describe inter-species interactions. We adopt fundamentals in “Shannon information” theory to understand variations in the amount of information in bits flow through the metabolic network of individual species models leading to reconstruction of “Compartmentalized community” multi-species and “Mixed-bag” models. Our study shows the “Compartmentalized community” model has the highest amount of information flow in most of the chemical composition combinations w.r.t biomass, secretion and uptake fluxes explaining the flow of information among both species and coexistence as a community. All tools applied for the metabolic model and subsequently combining them into community models of the soil microbiome were built into the DOE Systems Biology Knowledgebase (http://www.kbase.us), where they are available for use by the scientific community.

Bio-inspired computing via micro-sensors

Mohammad H Hasan, Mostafa Rafaie, Fadi Alsaleem

Microelectromechanical systems (MEMS) devices have shown great success as sensors in the Internet of things (IoT) age due to their favorable properties such as their small size and low power consumption. However, as IoT systems grow larger, processing data from these sensors in real time, using a central computing unit grows more demanding, especially when sampling data at high rates. In this work, we show the potential of MEMS devices as computational devices by shifting the computational burden in the system to a network from the central processor or the cloud to a network of MEMS sensors, interfaced to sensory-motor peripheries. We demonstrate this concept by performing a simple object characterization and tracking task via a MEMS network. The network is trained using a genetic algorithm scheme via the Holland Computing Center (HCC) supercomputer. After training, the network is shown to successfully track and capture circle objects and avoid line objects without the need for external digital computing devices. This concept may be expanded to produce smart MEMS devices, capable of sensing and computing within the same unit.

Predictive Model of Pancreatic Cancer Based on Radiomic Data Integration

Qian Du, M Baine, X Liang, A Kamal, and H Yu, D Zheng, C Zhang

Pancreatic cancer is extremely deadly. Historically, despite treatment, less than 20% of all patients were alive one year after diagnosis, and less than 5% at five years. Great patient and tumor heterogeneity exist. Not every patient responds to all therapies; considerable toxicity is associated with any therapy, even lethal to some patients; and treatment response or tolerability is not reliably predictable. Radiomics derived data, when combined with other pertinent data, could potentially produce accurate robust evidence-based clinical decision supporting systems, the outcrying need in pancreatic cancer care.

Uncovering the Damage Induced by Corrosion with 3D Simulations

Siavash Jafarzadeh, Ziguang Chen, Jiangming Zhao and Florin Bobaru

Synergistic effect of corrosive environment and mechanical stress may lead to a disastrous phenomenon called stress-corrosion cracking (SCC). SCC is a major cause of many fatal failures such as airplane crashes and collapse of bridges. We introduce a new 3D model for localized corrosion damage and SCC. Our model quantitatively predicts the experimentally observed damage pattern evolution in time. This model is a first-ever in simulating the synergistic effect of mechanics and corrosion along with realistic predictions of damage evolution in time.

Particle-in-Cell Simulations of Laser-Wakefield Accelerators

Junzhi Wang

Utilizing state-of-the-art powerful table-top laser systems, laser-wakefield accelerators (LWFA) provide compact and accessible means of electron acceleration, especially compared to conventional radio-frequency accelerators. Ultra-high intensity laser pulses are focused into a plasma to drive large amplitude plasma wakes where trapped electrons are constantly accelerated to near the speed of light over just millimeter distance. We present a numerical investigation of the physics underlining the LWFA process conducted using a two-dimensional, parallelized, relativistic particle-in-cell (PIC) simulation code EPOCH. Real plasma systems usually contain extremely large number of particles (order of 1014). In order to make the simulation feasible, the PIC code divides the simulation space into smaller cells containing a few macroparticles each representing a large number of real particles. The code then calculates the electromagnetic fields by solving Maxwell’s equations on the grid over a small time step, and pushes the macroparticles into new positions based on the calculated fields. However, the amount of macroparticles in a typical PIC simulation is still enormous (order of million), which often results in 100s GB of data. Due to the massive computational effort required, these simulations can be carried out only on supercomputers similar to the one at HCC. Capabilities to run these PIC simulations and interpret the results provide a crucial tool for modeling LWFA experiments and analyze experimental data obtained at the Extreme Light Laboratory at the University of Nebraska-Lincoln.

Bioinformatics and Proteomics Evaluations of Potential IgE Cross-Reactive Proteins in Novel Edible Insects and Shrimp

Mohamed Abdelmoteleb, Lee K. Palmer, Justin T. Marsh, Philip E. Johnson and Richard E. Goodman

RATIONALE: Historically, insects have been consumed in many countries, although rarely in Europe and U.S. The European Food Safety Authority recently recognized mealworm as a potential cross-reactive novel food with risk of allergic cross-reactivity for those with shrimp allergy. Regulators in the United States are asking developers to evaluate products containing cultured insects (crickets) for risks for crustacean allergic subjects, which we evaluated. METHODS: Transcriptomes of cricket, mealworm, silkworm, and locust and allergenic crustaceans were compiled using de novo and reference based assemblers. Transcripts were compared to allergens in AllergenOnline.org V18B using BLASTX, focusing on sequences of tropomyosin and arginine kinase. Abundance of mRNA of these proteins were estimated using RNA-seq quantification with RSEM software. Protein levels were measured using untargeted proteomics by LC-MS/MS to estimate the abundance of tropomyosin and arginine kinase. Potential IgE epitopes were predicted using five immunoinformatics programs. The predicted epitopes were compared to published epitopes from allergenic arthropod sequences. Specific IgE binding assays were performed to test IgE cross-reactivity using available shrimp and insect allergic sera. RESULTS: High sequence identities were found with high abundance transcripts of tropomyosin and arginine kinase compared to sequences of shrimp allergens. Proteomics confirmed the presence of isoforms. Direct binding and inhibition assays using protein extracts noted positive IgE binding for some allergic subjects. CONCLUSIONS: Some crustacean-allergic consumers are likely to experience cross-reactions if they consume foods containing proteins from crickets, mealworm, silkworm, or locust. Allergists should be aware and alert crustacean allergic patients.

Towards Improving transcriptome Assembly

Sairam Behera, Spencer Stream, Kushagra Kapil, Adam Voshall , and Etsuko N. Moriyama

Speeding HEP Analysis with ROOT Bulk I/O

Brian Bockelman, Zhe Zhang, Oksana Shadura

Distinct HEP workflows have distinct I/O needs; while ROOT I/O excels at serializing complex C++ objects common to reconstruction, analysis workflows typically have simpler objects and can sustain higher event rates. To meet these workflows, we have developed a “bulk I/O” interface, allowing multiple events’ data to be returned per library call. This reduces ROOT-related overheads and increases event rates -- orders-of-magnitude improvements are shown in microbenchmarks.

Unfortunately, this bulk interface is difficult to use as it requires users to identify when it is applicable and they still “think” in terms of events, not arrays of data. We have integrated the bulk I/O interface into the new RDataFrame analysis framework inside ROOT. As RDataFrame’s interface can provide improved type information, the framework itself can determine what data is readable via the bulk IO and automatically switch between interfaces. We demonstrate how this can improve event rates when reading analysis data formats, such as CMS’s NanoAOD.

MODIS-based gross primary productivity and GPP phenology for Arctic and boreal ecosystems

Gabriel Hmimina, Rong Yu, Karl Huemmrich, David Billesbach, Alexei Lyapustin, Yujie Wang, Zheng Xu, John Gamon

Arctic and boreal ecosystems are extremely vulnerable to global climate change. Understanding how these ecosystems respond to climate change is critical to global carbon budget estimation as well as the management of these vulnerable ecosystems. While eEddy covariance towers provide important carbon exchange measurements between ecosystems and the atmosphere with half-hourly resolution at the landscape scale, they cannot provide full coverage over the ABoVE domain. While optical satellites provide spatial coverage, they do not directly sample GPP, and often typically report “greenness” indices that can be confounded by various factors, particularly in northern ecosystems. For these reasons, we have developed a new satellite-driven GPP model integrating flux data and satellite MODIS MAIAC data within a light-response framework. The new MODIS MAIAC (Multi-Angle Implementation of Atmospheric Correction) dataset, which provides advanced atmospheric correction, and cloud detection, and surface reflectance measurements (including ocean bands over land) at sub-daily and 1km resolution. Using this newly available spectral information, we were able to estimateestimated photosynthetic capacity and limitation, potential GPP and the phenology of GPP. Our GPP estimation showed a es showed strong relationship with agreement with ground fluxeseddy covariance estimates, especially at daily/sub-daily scales (R2: 0.859), a weaker agreement (R2: 0.643) with GOME2 SIF product over 28 flux sites across the domain, and a good correlation with season length for the ABoVE domain. Over the study time span (2002-2014), GPP and its phenology across the ABoVE domain revealed clear spatial patterns of increase or decrease forin different geographic regions. In most cases, these trends were not statistically significant, partly due to the limited time span. However, model parameters related to photosynthetic capacity and limitation, based on the light response curve, provided additional information that might be useful in evaluating changing vegetation responses. Current plans are to further validate this GPP model and explore effects of disturbances (fire, drought, insect pests, etc.), succession, and changing phenology on annual GPP in these boreal and arctic ecosystems across the ABoVE domain.

Software Defined Radio based City-Wide Experimental Testbed for Next Generation Wireless Networks using Holland Computing Center Networks

Zhongyuan Zhao, Mehmet C. Vuran, David P. Young, Warren Humphrey, Steve Goddard, Garhan Attebury, Zahra Aref, Blake France, Baofeng Zhou and Mohammad M. R. Lunar

With the increase of wireless devices usage day-by-day, a robust platform for advanced testing and research also becomes crucial. A next generation advanced testbed is deployed to facilitate this necessity. In this testbed, a city-wide experimental platform is developed to facilitate the research of dynamic spectrum access, 5G, underground wireless communications, and Radio Frequency Machine-Learning in real-world environments. The full deployment is based on the collaboration of the university, city, and industrial partners. This Software Defined Radio (SDR) based testbed is composed of 5 cognitive radio and covers 1.1 square miles across two campuses of University of Nebraska-Lincoln and a public street in the city of Lincoln, NE. Each site is equipped with a 4x4 MIMO software-defined radio transceiver, USRP N310, with 20Gbps connectivity. Moreover, the street site contains an additional cognitive radio transceiver with underground 2x2 MIMO antenna. All these transceivers are capable to communicate at any sub-6GHz frequencies. The testbed has FPGAs at both the edge and and a cloud-based central unit, where data processing and storage take places. The cloud based central unit of this testbed is facilitated by Holland Computing Center (HCC).

HCC supercomputing cluster is used as a platform to access this testbed to facilitate education and researches in academic and industrial communities with a wide variety of wireless experiments. These experiments generate baseband signal data which initially stored at HCC storage. The baseband data then can be processed by high-performance HCC computing clusters depending on the requirements of the researcher. Currently, this testbed is used by 25 students for their course project of the course named Wireless Communication Networks. In this poster, detail deployment of testbed is presented. Also, some applications of this testbed are described with corresponding examples.

Calculations of Ultrasonic Properties for Simulated Microstructures Created Using DREAM.3D

Musa Norouzian, Showmic Islam, Nathanial Matz, Joseph Turner

Ultrasonic attenuation and diffuse scattering result from the interaction of ultrasound with the microstructure of polycrystalline samples. Researchers are now using these effects to quantify mean grain size with good success and progress is being made with respect to more complex grain morphologies and macroscopic texture. However, theoretical models of such microstructures can become untenable because the scattering theory requires the covariance of the elastic modulus which is an eighth-rank tensor. For this reason, computational models of polycrystals are often considered for which grain spatial statistics can be calculated directly. The influences of various microstructures are examined using such an approach with three-dimensional (3D) realizations of polycrystalline materials. Representative material volumes are simulated using DREAM.3D with multiple realizations. These realizations are then used to calculate the relevant grain statistics which are then used to determine ultrasonic properties. The results show a correlation between ultrasonic properties, particularly attenuation, and the simulated microstructures. The results are expected to aid in the development of simplified models that capture the dependence on microstructure properties.

Spontaneous self-assembly of Amyloid beta (1-40) dimer

Mohtadin Hashemi, Yuliang Zhang, Zhengjian Lv, Yuri L. Lyubchenko

According to the amyloid hypothesis, formation of soluble oligomers, and eventually fibrils, composed of Amyloid beta (Abeta) peptides in the brain is the etiological agent of Alzheimer’s disease (AD). While late stage causes of aggregation are somewhat understood, the molecular interactions that lead to the formation of fibrils from individual Abeta peptides are not well understood. It is challenging to investigate the interactions between monomers and low order oligomers at a molecular level by experimental approaches. Computational simulations, notably molecular dynamics simulations, enable us to examine the dynamics of the aggregation process, and to delineate the conformational transition from monomers to dimers. We have conducted in silico studies of Abeta(1-40) monomer interactions. One important observation from our long time-scale MD simulations is that the dimer does not have the extended beat-structures observed in fibrils. Rather, it remains largely unstructured and is stabilized by hydrophobic interaction and local Hydrogen bonding. The dimer structures were validated by comparing Monte Carlo pulling simulations with experimental data from single molecule AFM force spectroscopy. The rupture processes were in good agreement with our experimental data. The results suggest that the Abeta(1-40) aggregation is a complex process in which the monomer structure depends on the aggregate size.

Temporal Gene Coexpression Network Analysis Using A Low-rank plus Sparse Framework

Jinyu Li, Yutong Lai, Chi Zhang, Qi Zhang

Various gene network models with distinct physical nature have been widely used in biological studies. For temporal transcriptomic studies, the current dynamic models either ignore the temporal variation in the network structure or fail to scale up to a large number of genes due to severe computational bottlenecks and sample size limitation. On the other hand, correlation-based gene networks are more computationally more affordable, but have not been properly extended to gene expression time-course data. We propose Temporal Gene Coexpression Network (TGCN) for the transcriptomic time-course data. The mathematical nature of TGCN is the joint modeling of multiple covariance matrices across time points using a “low-rank plus sparse” framework, in which the network similarity across time points is explicitly modeled in the low-rank component. Using both simulations and a real data application, we showed that TGCN improved the covariance estimation loss and identified more robust and interpretable gene modules.