Research Projects

Research projects focus on the development and enhancement of algorithms, models, and software for applications in multiple domains that require high-performance computing resources. These domains include computational chemistry, biology, biomechanics, physics, astrophysics, mathematics, machine learning, materials science, earth science (seismology) and neuroscience.

Computational Modeling of the Ailing Heart

Mentors: Jason Bazil, Vani Malyala

This project will entail the use of an integrated myocardial energetics dynamical model to understand the impact of mitochondrial dysfunction on cardiomyocyte contractile performance, calcium handling, and free radical homeostasis. The model will be used to computationally test hypotheses generated to explain the experimentally observed decrease in cardiac function of diseased and damaged hearts.

Students working in this group should be familiar with undergraduate differential equations. Programming experience with MATLAB is useful but not required.

At the end of this project, participating students will (1) be able simulate and analyze systems of nonlinear ODEs using MATLAB; (2) gain exposure to numerical methods such as parameter optimization, sensitivity analysis, and uncertainty quantification; and (3) gain an appreciation for the physiochemical phenomena that govern cardiac metabolism, mechanical performance, and ion homeostasis.


Refining seismic structure of Los Angeles Basin using ambient noise adjoint tomography method

Mentor: Min Chen

The Los Angeles (LA) Basin, a sedimentary basin located in southern California, has been extensively studied due to its significance for oil production and concern for its seismic hazard to the high-rise buildings in the area. In order to accurately simulate seismic waves propagating through the LA Basin that cause damage during a large earthquake, we need a high-resolution and high-fidelity model to describe its subsurface seismic structure.

In this project, students and their mentor will work on a new data set from ambient noise interferometry and refine an existing three-dimensional model of the LA Basin. Adjoint tomography, a full waveform inversion based on a spectral-element method integrating high performance computing, will be used to iteratively update the model.

It is a bonus if the students applying to this project have taken coursework in Digital Signal Processing and Inverse Theory but these are not required. Experience with the Unix command line, Python, Matlab, and Fortran are strongly desired.

At the end of this project, participating students will be able to: (i) run basin-scale seismic wave simulations on a supercomputer using a spectral-element method based Fortran code; (ii) analyze, and visualize the simulation results using the Python-based toolkit, Seismic Analysis Code (SAC), and Generic Mapping Tools (GMT), which all are essential to Earth Science research. Furthermore, students will gain an understanding of inverse problem in seismic imaging and the impact of the strong seismic heterogeneities on observed seismic waveforms and therefore earthquake hazard assessment.


Streamlining Supernova Simulations

Mentors: Sean Couch, Luke Roberts, MacKenzie Warren, Carl Fields

Core-collapse supernovae are the explosive deaths of stars more massive than about ten times that of the Sun. Numerical simulations of the mechanism that drives these explosions are extremely complex, involving many different types of physics on a enormous range of scales. This project will involve exploring the sensitivity of core-collapse supernovae to variations in input physics. Students will carry out simulation parameter studies and analyze resulting data, including producing relevant visualizations.This project will involve the development and use of a simulation management and analysis software package for supernova simulations on many different computer systems ranging from laptops to the world's largest supercomputers. This tool will be used in executing the supernova simulations to be used to better understand critical physical sensitivities in the process that drives these stellar explosions.

Students working on this project will use python, bash, a little Fortran.

At the end of this project, students should be very familiar with software version control, the use of common Python tools for computational science and data analysis, and the efficient planning and execution of numerical simulation. Students will also be introduced to the physics and mathematics of astrophysical simulation.


Simulation of the effect of a novel kinase inhibitor

Mentors: Alex Dickson, Arzu Uyar

Kinases play a pivotal role in the modulation of many cellular processes, and their malfunction is implicated in a variety of diseases including a majority of human cancers. Beginning with a structure solved by an experimental collaborator, this project will involve screening a large database for potential "hits", as well as simulation of the ligand-protein complex to identify stable states, as well as ligand-binding transition states, that can lead us to more potent ligands that can be tested experimentally.

Students working on this project must be familiar with introductory chemistry. Knowledge of undergraduate physical chemistry and biochemistry is useful but not required. Prior experience programming with python is helpful but not required.

At the end of this project, students will be able to run virtual screening to find potential new drugs, and run molecular simulations to generate trajectories of molecular motion.


Multi-scale modeling of cardiac fluid-structure-growth

Mentors: Tong Gao, Lik Chuan Lee

This project aims to develop a strongly-coupled cardiac electromechanics-fluid-growth computational modeling framework that takes into account some key physics occurring in the beating heart, including: cellular excitation-contraction coupling processes, contribution of constituents found in the cardiac tissue, and macroscale fluid-structure interactions between the heart wall and blood.

Students working on this project will (1) gain basic knowledge of cardiovascular mechanics, (2) learn to use basic computation software/packages in cardiovascular modeling and simulation, (3) be able to perform engineering analysis to interpret numerical data and make connections to some abnormal mechanical behaviors.


Machine learning from Quantum Computing

Mentors: Matthew Hirn, Yue Qi

Machine learning algorithms are obtaining state of the art results in computer vision, natural language processing, and auditory signal processing, amongst other areas. Recently, there has been a push to utilize machine learning to solve problems from chemistry and physics, by drastically speeding up complex computations arising in quantum chemistry. In this project, students will develop cutting edge machine learning algorithms to solve problems in materials science. Such problems may include lithiation of Si electrode and fracture of electrode materials. Students should have an interest in either machine learning or materials science (or both!), and a desire to take on projects at the interface between these two fields.

Students working on this project must be familiar with linear algebra. Knowledge of MATLAB and/or Python is useful but not required.

Developing machine learning approaches to infer context-specific networks of genes

Mentors: Arjun Krishnan, Jianrong Wang

Hundreds of types of cells in our body - in the heart, kidney or brain - are able to perform sophisticated tasks because genes in each of these cells work together in the form of a complex network. The network of genes is also critical in how diseases like cancer or autism arise and affect us. Fortunately, we have been accumulating an enormous amount of data about human genes in tens of thousands of conditions. In this project, students will develop machine learning approaches for integrating these big data sets to predict gene networks in diverse biological contexts.

It is a bonus if students: (i) are familiar with introductory machine learning and statistical modeling concepts, (ii) have practical experience with machine learning toolboxes and/or analyzing large data sets but these are not required. Prior knowledge of MATLAB or R is useful but not required.

At the end of this project, participating students will be able to pre-process data, run multivariable statistical analysis and machine-learning using a Python/R-based toolkit. Moreover, students will have learned fundamental concepts in machine learning, statistical inference, and high-performance computing on biological big-data, along with an introduction to modern problems in genomics, systems biology, and precision medicine.


Effective Potentials for Simulating Non-Equilibrium Quantum Many-Body Systems

Mentors: Michael Murillo, Yongjun Choi, John Luginsland

Solving the time-dependent Schrodinger equations for thousands of particles remains extremely expensive. Rather than attack the problem directly, we will develop semi-empirical potentials motivated by the Schrodinger equation but simplified enough to allow for large scale computations. Errors made are mitigated through training sets from empirical data. The resulting model will be implemented in our group's pure Python MD code Sarkas and applied to problems of current interest.

Students applying to this project must be familiar with quantum mechanics at some level. Prior knowledge of Python with Numba is also useful but not required.


Implementation of dynamic network algorithms in a highly parallel architecture

Mentors: Mark Reimers, Michael Moore

Students will work with grad students and post-docs developing algorithms for analyzing optical imaging of neural activity, specifically during the processes of encoding and recalling memories in rodents and people. The algorithms will attempt to: (i) identify recurrent firing patterns of many neurons over hundreds of thousands of image frames; (ii) to infer strong connections within the brain. Students working on this project will develop highly parallel implementations of these algorithms for use on the MSU HPCC, and possibly on EC2.

Familiarity with parallel computing is useful but not required. Experience programming with MATLAB is also useful but not required.

Students working on this project will learn how to work with large-scale dynamic brain imaging data, and how to identify closer network connections using parallel algorithms.


Adaptive mesh refinement for the smoothed boundary method

Mentor: Hui-Chia Yu

The smooth-boundary method (SBM) is a numerical method that is powerful for solving differential equations in complex geometries. This project aims to develop an algorithm for mesh refinement that will be applied to the framework of SBM. With adaptive mesh, the accuracy and efficiency of SBM is expected to greatly improved.

Experience programming with Matlab and Fortran will be helpful, but is not required.

At the end of the project, participating students will gain experience developing SBM code and simulating materials phenomena in complex microstructures with adaptive mesh refinement. These skills are particularly useful for analyzing complex phenomena occurring in energy materials.