Winners of Neukom Graduate Fellowships have been announced for the 2013-2014 academic year. Fellowships will provide a full year of funding, including stipend and benefits, to Ph.D. students engaged in faculty-advised research in the development of novel computational techniques as well as the application of computational methods to problems in the Sciences, Social Sciences, Humanities, and the Arts.
The 2013-2014 winners are:
How stars form is the backbone of many astronomical studies, from galactic formation and evolution to cosmology. Galaxy clusters, gravitationally-bound groups of hundreds of galaxies to over a thousand, are an excellent laboratory for studying how the interplay between galaxies, intergalactic gas, and dark matter affects the formation of stars.
In general, clusters are dominated by the oldest, largest galaxies in the Universe that exhibit much lower rates of star formation (SF) than their counterparts in less dense environments. Typically, clusters' only star-forming galaxies are found on the cluster outskirts before their interaction with the clusters truncates their SF. This segregation between passive and active galaxies is typical of "relaxed" clusters. However, recent studies confirm that most galaxy clusters are not relaxed: their galaxy distributions show complicated substructure, indicative of an active past of smaller clusters merging together. The canonical picture of cluster centers dominated by passive galaxies thus no longer applies, and it is possible for cluster mergers to re-ignite SF in normally quiescent environments.
This project will analyze a large number of galaxy clusters in various states of merging to determine whether a correlation exists between cluster interaction and SF. Our research will use a collected sample of over 120 galaxy clusters containing over 10,000 galaxies using public optical and spectroscopic data from the Sloan Digital Sky Survey, public infrared data from the Wide Field Infrared Survey Explorer, and supplemental data from the literature. Analyzing such a large number of clusters and galaxies is only tractable using statistical and computational techniques, which will help determine whether a correlation exists between cluster merging and SF.
Using these methods, we will determine where, when, and with what strengths SF is occurring in the large sample of clusters. Furthermore, by employing several techniques to measure SF for each galaxy, we can investigate any potential systematic biases of each method.
This study makes use of recent computational results that have used one-, two-, and three- dimensional tests for substructure. Using these statistical results, we will measure the strength and position of galaxies' SF as functions of the density of substructure.
The complex interactions at play in galaxy clusters makes them inherently diverse, which introduces scatter into correlations between measurable quantities. Therefore, detecting a correlation between cluster mergers and SF must involve robust statistical analysis. In analyzing this quantity of galaxy clusters and galaxies, our project will be statistically rigorous in a way no other study has been. Determining how stars form in galaxy clusters provides broad insight into how a wide variety of gravitational and hydrodynamical mechanisms affect the growth of galaxies.
Gene expression produces the complex machinery necessary for cellular life, and its regulation is a crucial means by which cells can assume specific functions. One of the cell's implicit constructs that accomplishes this regulation is its network of gene-gene interactions, in which the product of one gene directly influences the expression of another gene, as happens with transcription factors (TFs). This network is referred to as a gene regulatory network (GRN). Because GRNs are often involved in crucial biological functions, it is important that they are robust to genetic perturbation, such as gene knock-out (Jeong et al., 2001) or the rewiring of regulatory interactions (Isalan et al., 2008). Several theoretical studies have attempted to elucidate the source of this robustness, and one source appears to be GRN topology.
Assortativity is a topological property that has been shown to vary in real-world networks (Newman, 2002; Foster et al., 2010). The assortativity of a GRN measures the tendency for genes to interact with other similar genes. Our recent work led us to further inquiry: Do biological GRNs display evidence of nonrandom degree assortativity, and does this contribute to their robustness?
This study will address the question of whether the assortativity of biological GRNs appears to contribute to their robustness by investigating a recently established dataset of GRNs from a diverse set of human cell lines and tissues types (Neph et al., 2012b). An examination of the 3-node network motifs in these GRNs suggests that they share similarities with each other and with other directed biological networks (Neph et al., 2012a). However, their assortativity has not been studied, and it remains to be seen what insight their global network properties can provide about their robustness.
For directed networks, there are four different kinds of assortativity depending on whether the in- or out-degree is considered for each node in a pair of nodes that share an edge. These are referred to as out/in, in/out, out/out, and in/in; our previous work focused exclusively on the out/out case. Data reveal that nearly all the GRNs possess high out/out assortativity, which is precisely what our theoretical results have indicated is a potential source of robustness. It also appears that out/in assortativity is very high, whereas in/out and in/in appear close to random. This represents the first time such an approach has been applied to biological GRNs.
Our research will use GRN models that resemble the biological GRNs with respect to all four forms of assortativity and will determine whether GRNs with the biological assortativity "signature" appear more robust than random GRNs.
This will indicate whether the increased out/out assortativity is capable of producing highly robust GRNs in the context of the rest of the assortativity signature. Further results will reveal how the different forms of assortativity interact in order to generate robustness and whether the observed biological signature is optimized for robustness. This work will potentially lead to a greater understanding of how natural processes shape the topology of biological networks.
Magnetism has fascinated scientists and laypeople alike for countless centuries. 2,500 years ago, the attractive forces of iron were first described and it was another 1,700 years before the compass was invented. Today, magnetism remains one of the most applicable fundamental properties of nature, providing the technology for generators, computer memory, speakers, MRI, car engines, and more.
Increasingly smaller magnets are now possible, and so-called magnetic nanoparticles (MNPs) with diameters ranging from 1 to 100nm, have garnered the attention of many fields of science. Particularly in biomedical physics, because of their comparable size, they are fascinating tools capable of probing the fundamental workings of cells and molecules. MNPs are also extremely applicable in medicine. They can be used as drug delivery tools and therapeutic as well as contrast or imaging agents. But, to advance these applications, a detailed understanding of MNP behavior is required.
This project will separate out the signatures of multiple dynamic mechanisms, from magnetic alignment to viscous or Brownian relaxation to equilibrium, and, looking forward, to chemical binding, fluid flow, and internal magnetic rearrangement.
To model nanoparticles, a typical dynamical model is insufficient. Due to the constant collisions of the particles with the even smaller molecules of the liquid they are suspended in, exact predictions are fruitless. Instead, in the same vein as others, we will employ a stochastic Langevin equation to describe the time evolution of random variables.
A comprehensive benchmarking study was completed recently that verifies the model predicts known theoretical and experimental results. We are beginning the exciting phase where confidence in the model allows us to make predictions about novel physics. We hope by analyzing subtle changes encoded in the magnetizations, signatures for respective mechanisms can be developed.
Many of the applications of MNP involve biological interaction. Thus, we are studying the addition of a biochemical binding term to the differential equation. Preliminary simulations demonstrate variation in magnetization dynamics as well as interesting oscillation signatures in directions perpendicular to the oscillating magnetic field. This may allow statements to be made about the amount or strength of binding that a nanoparticle experiences when faced with immune system response, potentially informing cancer detection methods or other biological sensing capabilities.
Overall, studying the rotations of nanoparticles, given constraints like viscosity, temperature, and especially binding, is informative to many different fields of science. Insight that is gained through theoretical modeling could be used importantly in nanoparticle sensing – notably for cancer detection – and also in therapy or imaging.
The human brain consists of distinct regions, which interact with one another to support and execute complex cognitive functions. Understanding how the brain processes sensory inputs using a multitude of interconnected cortical subsystems remains a major challenge in neuroimaging research.
In the physiology of brain function, neural activity is accompanied by cerebral blood flow changes and the dynamical relationship between the two is termed neurovascular coupling. This coupling between neural activity and cerebral blood flow is mediated by the neurovascular unit, which includes neurons, capillaries, astrocytes, and microglia that work together as a complex system. Dysfunction of the neurovascular unit has been suggested as a causative factor in neurodegenerative disorders such as Alzheimer's disease, migraine and epilepsy, and hypertension and ischemic stroke.
The purpose of this project is to quantify the functional relationships between different brain regions by taking into account the biophysical process associated with neural activity. Our research proposes to establish new methods for studying the neurovascular coupling relationship from multimodal neuroimaging experiments. After identifying a robust mathematical model, we will map the neurovascular coupling relationship at the network level. These networks will provide insight into the organization and structure of neural and hemodynamic correlates, which may change dramatically in a diseased brain. Using computational methods in graph theory, we will then assess the stability of the network topology in healthy subjects and patients with neurovascular disease.
A dynamical system model will be used to capture the causality between neural activity and the hemodynamic response. This novel computational aspect can become a platform for studying human brain function by researchers in Psychological and Brain Sciences and physicians in Neurology. There is enormous potential for studying disease mechanisms in the brain from neurovascular disorders like stroke, epilepsy, and AD.
The aim of this work is to help advance neuroimaging research by developing a data analysis method that reveals fundamental new understanding of the dynamic relationship between neural activity and cerebral hemodynamics. This work will also lay the foundations for studying human brain function in the context of neurovascular coupling networks and will serve as a novel computational tool in neuroinformatics.
Last Updated: 9/4/15