The Phase Diagram of QCD
The long-term goal of this project is to determine from first principles the properties of quantum chromodynamics (QCD) as a function of the temperature and the densities of the u, d, and s quarks. A similar effort from experimental physicists is under way at particle accelerators RHIC and LHC, by colliding heavy ions. Theoretical progress is hampered by the “sign problem”: the fermion determinant becomes complex when the quark density (or equivalently the quark chemical potential) is non-zero, which makes the usual Monte Carlo sampling impossible. These researchers have been pursuing, very successfully, the strategy of making the chemical potential mu imaginary. Positivity of the determinant is restored, and standard Monte Carlo can be used. The imaginary-mu results can be analytically continued to real mu. Moreover, the critical points found at imaginary mu imply scaling laws extending to real mu. Thus, the researchers want to constrain the real-mu properties, by pinning down the phase diagram for imaginary-mu, accumulating more knowledge by letting the quark masses take arbitrary values. These simulations of lattice QCD with imaginary chemical potential are quite standard. The researchers are determining the critical surfaces and tricritical lines in the extended phase diagram by well-established finite-size scaling techniques. Computing resources are absorbed in the simulation of large systems near criticality.
Algorithms and Analysis for Natural Systems
These researchers are involved in two projects requiring large-scale computation. Both projects are developing new algorithms for understanding natural systems and are applying those algorithms to prominent real-world problems.
In the first project, the researchers are continuing their development work on Order-1 algorithms for large-scale, individual-based, equation-free modeling in ecology, epidemiology, and economics, and applying those agorithms to a whole-nation test case that addresses the unexpected resurgence of tuberculosis in recent decades. These Order-1 algorithms cover time and group management and run at a fixed speed regardless of whether the simulation encompasses sixty individuals or sixty-million. That allows the researchers to extend their investigations beyond what has normally been possible.
The second project is in its early stages. The researchers are developing parallel algorithms for processing fine-scale elevation data of the earth’s surface in order to identify new watershed configurations. They are applying LIDAR data at one-square-meter resolution in a regional test case to learn how to decouple pollutant-laden waters in the artificial watershed of drain tiles and ditches from fresh waters in the natural watershed of lakes, ponds, rivers, and streams.
Modeling Countercurrent Shear in Practical Devices
These researchers are extending their previous MSI-supported work that looked into adding a counterblowing device to a backward facing step. The simulations to date modeled flow over a backward facing step using ANSYS CFX under isothermal conditions. At the sudden expansion a small device was added through which air was blown in opposition to the primary stream. The additional shear was sufficient to activate a global instability and dramatically alter the mixing of the well-established backward facing step. Backward facing step flow is a geometry used to model a RAM jet combustor and an active area of research by our group and others. The simulations and experiments to date have shown counterblowing to be a good candidate to increase burning rates. Experiments are underway to quantify the degree to which the blowing can improve combustion and provide benchmark data. The researchers are simulating the counterblowing backward facing step with combustion added. This work builds on existing computer simulations and continues to be two-dimensional and use both laminar and turbulence models. The researchers will initially use simple combustion models and add complexity as time and resources allow.
Population Pharmacokinetic Modeling of Anticancer and Other Therapeutic Agents
This population pharmacokinetic modeling project is being conducted in support of completed and ongoing clinical trials conducted by the Mayo Clinic Comprehensive Cancer Center. Pharmacokinetic data obtained from preclinical and Phase 0–III studies are utilized to develop comprehensive compartmental models that describe the pharmacokinetic behavior of investigational drugs and their metabolites. MSI resources are used for population-based analyses of complex models and large datasets as the underlying calculations are computationally intensive and require the simultaneous use of several processors. The population-based models are developed through the use of the software program NONMEM and are utilized to quantify the pharmacokinetic variability exhibited between study subjects. Depending on the available data, correlations between pharmacokinetics and pharmacodynamic outcomes will also be investigated. Taken as a whole, the intent of this investigation is to utilize the population pharmacokinetic modeling approach to ascertain the influence of clinically relevant and measurable factors on the pharmacokinetics and pharmacodynamics of anticancer and other therapeutic agents with the ultimate goal of improving drug dosing guidelines on the individual patient level.
Pediatric Hemiplegia: Synergistic Treatment Using RTMS and CIT
Paralysis following stroke stems not only from the loss of neurons killed by the stroke but also from the loss of neurons lying dormant in the stroke hemisphere. One of the reasons viable neurons become dormant (down-regulated) is because of excessive interhemispheric inhibition imposed on them from the nonstroke hemisphere. Suppression of the source of this excessive interhemispheric inhibition can be achieved with the noninvasive method called repetitive transcranial magnetic stimulation (rTMS).
The specific aims of this study are to determine the efficacy, mechanism, and safety of a series of five treatments of 6-Hz primed low-frequency rTMS applied to nonstroke hemisphere and combined with motor learning training to promote recovery of the paretic hand. Forty subjects with stroke will be randomly assigned to one of four treatment groups; rTMS-only, Track-only, rTMS-sham, and rTMS-combined. The hypotheses are: the rTMS/combined group will show the greatest improvements in hand function; and the rTMS/combined group will show the greatest improvements in cortical excitability using paired-pulse TMS testing and in brain reorganization measured with resting state functional MRI (rfMRI) and diffusion tensor imaging (DTI). Computationally intensive tasks, including pre- and post-processing of these imaging data, require the supercomputers in order to progress the work in a reasonable time frame.
Return to this PI's main page.
Genomics, Metagenomics, and Transcriptomics of Fungal Pathogens of Invertebrates
Using a combination of next generation sequencing, phylogenomics, genetics, and natural products chemistry, the Bushley lab examines the evolution, diversity, and functions of fungal secondary metabolites. Current research projects utilizing MSI resources include:
- A comparative population genomic study of the evolution of NRPS secondary metabolites among strains of the beetle pathogen and cyclosporin producing fungus Tolypocladium inflatum using PacBio sequencing
- A comparative genomic and transcriptomic analysis of interactions of insect pathogenic and endophytic fungi with both plant and insect hosts
- A metagenomics study of fungal pathogens of the soybean cyst nematodes to elucidate patterns of distribution in natural and agricultural ecosystems and potential roles in mediating resistance to nematodes
- Metagenomics analyses of tropical endophytic fungi of Papua New Guinea and potential anti-herbivore and anti-cancer activity.
These research projects utilize HPC computing for de-novo genome sequencing and assembly, RNA-Seq, network analysis, large-scale phylogenomic analyses, and population genotyping.
Return to this PI's main page.
Discover Evidence of Personalized Health Management to Improve Healthcare Outcomes
For most medical problems, clinical (patient) heterogeneity influences treatment efficacy and results in variations in outcome in one-treatment-fits-all settings. However, an opportunity exists to improve outcomes while reducing costs using currently existing treatments when we understand how clinical heterogeneity influences treatment efficacy and how much of a difference exists among treatment options. Knowledge of how to personalize treatment to account for this clinical heterogeneity is the key to optimizing outcome and improving treatment efficiency.
Projects by this research group use this opportunity to improve healthcare outcomes, whose evidence is extracted from electronic health records. These projects are:
- Mining personalized Alzheimer's Disease treatment from data
- Predicting a cognitive decline curve for Alzheimer disease
- Using a data-mining approach to facilitate efficient use of nursing resources
In general, each project derives evidence of improved-outcome evidence associated with treatment options, patient characteristics, and interactions. They benefit largely from MSI computing resources.
Return to this PI's main page.
Enzymology and Biotechnology of Protein Prenylation
Protein prenylation is an irreversible covalent post-translational modification found in all eukaryotic cells, comprising of farnesylation and geranylgeranylation. Three prenyltransferase enzymes catalyze this modification. This three-step process increases protein hydrophobicity, and often leads to plasma membrane association. Prenylation serves as the first critical step for membrane targeting and binding, as well as mediating protein-protein interactions of a large number of Ras proteins; heterotrimeric G-proteins also require prenylation for activity.Significant interest in studying protein prenylation originally stemmed from the finding that this modification was necessary to maintain malignant activity of oncogenic Ras proteins although now it is known that prenylation is important in a wide ranges of diseases. These researchers are using computer-based methods for three subprojects within this area. They include:
- Design of caging groups used to mask the activity of substrates and inhibitors of protein prenylation
- Bioinformatic analysis of proteomic data obtained using probes that allow selective detection of prenylated proteins
- Modeling of prenyltransferase structures to design mutations that alter substrate specificity
Return to this PI’s main page.