Student Posters Repository

In the present "RNA World”, Ribonucleic acid (RNA) plays a vital role beyond translation of information from DNA to proteins as depicted in the central dogma of biology. In has been elucidated that various forms of non-coding RNA (ncRNA) are involved in different cellular processes from catalysis in protein synthesis to gene silencing. RNA is also associated with several diseases, so understanding the RNA structure - function is of vital importance. Determining the structure of the RNA in the presence of drug-like molecules is a crucial step in any drug development campaign.  As such, there is a need for the development of fast, easy, and "precise prediction" methods for determining the 3D structure of RNAs. Chemical shifts have become an important tool for the Structure-Activity Relationship (SAR). Therefore our research is focused on developing models, using this structural information, with better accuracy, that, in turn, could elucidate how the key structural features of RNA systems are associated with its function, how systems interacts, what they bind to and design drug agents that exploit these structural features to solve medical problems. 

An Accurate Coarse-Grained Model for Polysaccharides in Solution and Close to Interfaces

Levan Tsereteli, Dr. Andrea Grafmueller

                      Max Planck Institute of Colloids and Interfaces, Am Muhlenberg 1, Potsdam-Golm, Germany

  Levan.tsereteli@mpikg.mpg.de

Computational models can provide detailed information about molecular conformations and interactions in solution, which is currently unachievable for experimental techniques. Here we provide efficient and precise model for studying conformational properties of long chain polysaccharides in water solution, under different physico-chemical conditions (pH, ionic strength) as well as interactions with various surfaces (Lipid Bilayers, Vesicles, Silica Nano particles etc.). Our model is validated against experimental data for Chitosan polysaccharides with various degree of deacetylation and is applicable to any polysaccharide in a good solvent.

Within our work we developed Coarse Grained (CG) Force Field (FF) with bottom-up approach [1]. Multiscaling was based on Molecular Dynamic (MD) and Metadynamic simulations performed with Amber Glycam All-Atom (AA) FF for the polysaccharides. Pivot Move (PM) Monte Carlo (MC) simulations of CG polysaccharides were used for characterizing the properties of the sugar polymers in implicitly solvent. All available experimental data could be closely reproduced.

We were able to characterize microscopic and mesoscopic structural properties of the long polysaccharide molecules at various pH and ionic strength of the solvent. Consider cooperative impact of the relative stiffness of the particular glycosidic angles and steric interactions. Determine protonation rate and pKapp of the polymer, radius of gyration, persistence length and characteristic ratio.

After the successful characterization of the polymers in good solvent, studies of their interactions with various surfaces of interests were performed. Fully titratable models for the interactions with liposomes with various DOPG/DOPC ratio was employed.

Adsorption on surfaces of various metal oxides was studied with constant charge density of the surface. Currently we are able to simulate systems as dense as composed out of 10 vesicles and 100 polymers each of 1000 monomer unit length. Considering the system size the model achieves a very high level of chemical accuracy.

Simulations were performed with a code based on the cross platform .Net framework Mono and OpenTK libraries. Simulations of a polymer with 1000 monomers including all force field contributions took 24 h for 5000000 PM steps for Chitosan with 90% DD on a Intel(R) Core(TM) i7-3770 CPU 3.40GHz processor.

References:

[1] An accurate Coarse-Grained Model for Chitosan Polysaccharides in Solution. Levan Tsereteli, Dr. Andrea Grafmueller. Plos One. Submitted.

To further the ability to design membranes for separation/fractionation of deformable particles (such as, cells, liposomes, vesicles, and droplets in emulsions and oil-water suspensions), we have developed a 2-d multiscale computational approach to study how the pressure drops and bulk flow within the depth of a porous "membrane" influences the mobility of an immiscible droplet through that structure. We use a combination of the extended finite element method to describe the creeping fluid flow (Re ~ 0) inside a portion of a filtration membrane with an embedded fluid droplet, coupled with a particle method that interpolates the droplet's interfacial position, as well as, the corresponding velocity and pressure fields using least square fitting. We calculated how the combination of several model 2-d porous network domain geometries (pore size and distribution), and a soft particle's deformation-related property (surface tension), influences the particles' velocity relative to the bulk fluid flux (aka sieving) in model porous domains made up of circular obstacles. The focus in this paper is on the scaling relationship between the particle's properties, the geometry of the system, and the overall droplet's sieving through a periodic domain. In this work we present first the case of a droplet permeating through an individual pore to determine its critical pressure. Then, the base case of a single pore and droplet is extended to include arrays of obstacles (creating a porous network domain) with different droplet volume fraction. These cases can provide a set of scaling rules to guide membrane design for droplet separation purposes. 

HARVEY is a massively parallel hemodynamics simulator that is based on the lattice Boltzmann Method (LBM) developed by our lab. LBM is an attractive computational method to simulate fluid flow due to its ability to work with complex geometries and reliance on nearest neighbor communication. I am currently working to extend its capability to simulate high Reynolds number flow using a newly designed boundary conditions. My role is to implement the boundary conditions efficiently in HARVEY by not affecting the runtime performance or create a memory burden. This is done by an encoding scheme to compress the different boundary conditions and the front-loading of the work for the boundary conditions that allows the computations in the simulation to be reduced to constant time operations. Our approach was tested on the lid driven cavity benchmark problem, which resulted in achieving strong and weak parallel scaling. To test the full suite of the boundary conditions, we intend to test it on fluid flow around an object. This is a computational complex problem necessitating the efficient use of HPC due to capturing the vortex formation or shedding for a given resolution.

To provide firefighters with more information about the evolving conditions during a structural fire event, a new computing infrastructure is under development at the University of Michigan. Through the use of existing and novel software packages, this recent effort provides a new link between the scene of a fire event and the responding fire department. Specifically, it is envisioned that buildings will someday be instrumented with sensors that monitor key fire signatures such as gas temperatures and gas-species concentrations. The live stream of data would be transmitted wirelessly to a computing workstation that would interpret the fire data and communicate actionable information to firefighters via a building information model (BIM). Such a system would provide a better understanding of the conditions within a building and will enable informed decision-making by the fire department that could lead to fewer deaths and property losses in structure fires. This presentation will showcase the current stage of development of this system by using the new software with simulated fire data in order to highlight some of the computational and visualization features that may someday be coupled with a live stream of fire data originating from an advanced sensor network.

Direct analysis of radiation emission in PIC codes is limited by grid resolution. In certain laser-plasma interaction scenarios with relativistic particles this is particularly challenging due to existence of very different spatial/time scales. As an example, betatron radiation in laser-wakefield accelerators can easily reach the X-ray regime (0.1 - 10 nm wavelength), while the laser wavelength is on the order of the micron and the plasma wavelength in the range of 10 to 100 microns (for plasma densities of 10^17 to 10^19 cm^-3). The calculation of radiation emission from the particle trajectories using the emissivity formula from classical electrodynamics has been widely used as an alternative. In this work, a parallelized numerical diagnostic code for radiation emission modeling using this method is presented. Its parallelization is based on MPI and the input/output relies on the HDF5 library. The code features (2D energy and 1/2/3D spectrum diagnostics, polarization and orbital angular momentum diagnostics), structure and possible improvements are discussed and examples of applications to laser-plasma interaction scenarios are presented.

Algebraic Multigrid is a popular method to approximate the solution to linear systems of equations. There are some challenges in implementing different parts of the solver, such as sparse matrix-vector multiplication and matrix-matrix multiplication. A simple heuristic to do these operations in parallel will be explained.

Flame retardant/polymer testing is done with test bars subjected to heat or flame in tests such as the European Union glow-wire test and the Underwriters Laboratory 94 methane flame test. A conventional extruded polymer test bar is a homogenous blend of flame retardant and high impact polystyrene; the 12 wt% flame retardant sample passes the UL 94 test while the 3 wt% sample fails.The glow-wire test was adapted for the CAMD tomography/interferometry beamline to study the effect of heat on plastic polymers containing different amounts of flame retardant. Stepped grating X-ray interferometry was performed as a function of brief heating episodes to observe the evolution of the absorption, dark-field (scattering) and differential phase contrast images. The heater was positioned on a counterweight-trolley-rail system ensuring constant 1 Newton force on the deformable sample. This setup allowed automated, repetitive near-burning and interferometry 2D imaging. Post-heating, samples were selected for interferometry/tomography. Numerical simulations were also conducted using COMSOL, to investigate the evolution of bubbles and the temperature gradient across the heated polymer