2020 RESEARCH SEED AWARDS
Petra Terraces Archaeological Project
The Petra Terraces Archaeological Project (PTAP) aims to study the long-term history of agricultural infrastructure in the hinterlands of the ancient city of Petra in southern Jordan. The project brings together an international team of archaeologists, anthropologists, geologists, and architects to study the construction, use, repair, and collapse of ancient terrace walls, dams, and related anthropogenic features that have enabled agriculture in a semi-arid environment over the last three thousand years. PTAP’s purpose is to produce a detailed, diachronic analysis of how people have shaped the local landscape by controlling—and at times also failing to control—flows of water and sediments along a single major watershed north of the city. The project builds on previous work by Brown archaeologists, both in the Petra city center, and more recently, in Petra’s outskirts. Specifically, it expands and refines the findings of an ambitious regional survey (based at Brown) that documented the ubiquity of anthropogenic modifications in the northern hinterlands of the city. PTAP will shed light on various matters of urgent importance in the contemporary Levant such as environmental inequality, human resilience and adaptability, and local responses to colonialism and imperialism, while also strengthening Brown’s ties to academic and non-academic communities in Jordan and neighboring regions.
PI: Felipe Rojas, Associate Professor of Archaeology and the Ancient World and Egyptology and Assyriology
Finding Social Narratives in Big Data
Decades of social science research has taught us that social discourse is governed by narratives, stories about how actors’ intentions and actions relate to outcomes. The narratives one buys into reflect one’s affiliations and how one makes sense of the social world. We are a team of cognitive scientists, applied mathematicians, and computer scientists. We propose to use cutting edge machine learning technology—OpenGPT-2, a deep learning algorithm partly developed at Brown—to discover the narratives that guide social discourse using big data from online sources. In the process, we hope to develop a general theory of narrative that can be computationally realized. Our plan is to examine text generated by OpenGPT-2 on three classes of topics: Topics showing significant polarization in the US (e.g., immigration); topics with less polarization, reflecting emerging consensus (e.g., gay marriage); and non-political topics such as video game strategies. We will analyze the text with the goal of developing a formal model of the narratives underlying discourse, apply that model to enhance understanding of what OpenGPT-2 is doing and to improve the theory and measurement of narratives in discourse, and increase the scalability of these methods to ease application to different timeframes and disparate cultures. The work is inherently interdisciplinary involving both computational and social sciences, with implications for machine learning and theories of narrative. It will cement Brown’s status as a hub for cutting edge research that applies computational methods informed by cognitive science to social issues.
PI: Steven Sloman, Professor of Cognitive, Linguistic and Psychological Sciences
Co-PIs: Bjorn Sandstede, Professor of Applied Mathematics, Director of the Data Science Initiative; Ellie Pavlick, Assistant Professor of Computer Science; Carsten Eickhoff, Assistant Professor of Medical Science, Assistant Professor of Computer Science.
Investigating the neural basis of sequential control in obsessive compulsive disorder
Obsessive-compulsive disorder (OCD) is a neuropsychiatric disorder that affects ~2% of the population over their lifetime and is associated with impairments in behavioral flexibility and compulsive, ritualistic behaviors. Such behaviors can often be conceptualized as sequences that are stuck in a “loop” and have to be performed repeatedly and in a ritualized way, such as hand washing or lock checking. As shown by Dr. Desrochers, control over behavioral sequences is governed by an increase in activity (“ramp”) from the beginning to the end of the sequence in the frontal polar cortex (FP) of healthy adults. Sequence performance is disrupted with transcranial magnetic stimulation (TMS) to the FP. The FP is also known to show reduced recruitment in OCD during a variety of experimental paradigms. We hypothesize that patients with OCD may have a deficit in this ramping activity in the FP. Dr. Garnaat is a licensed clinical psychologist and expert in OCD who has recently been awarded a grant to study cognitive flexibility in OCD patients (and healthy age-matched controls) using fMRI and TMS, with the FP as a main target. We therefore propose to expand Dr. Desrochers’ research into a new field, research in clinical populations with OCD, in collaboration with Dr. Garnaat. The proposed project will capitalize on the patient recruitment, assessment, and experiments already scheduled with Dr. Garnaat’s funded project to add a component to scan OCD patients performing a sequential task. These studies will compliment both lines of research, and contribute to our fundamental understanding of the neural circuits underlying OCD.
PI: Theresa Desrochers, Rosenberg Family Assistant Professor of Brain Science, Assistant Professor of Psychiatry and Human Behavior
Co-PI: Sarah Garnaat, Assistant Professor of Psychiatry and Human Behavior (Research)
New biomarkers for neurodegenerative diseases
Neurodegenerative diseases represent a major threat to human health. We here propose to establish the experimental protocols and analyses pipelines for the discovery of new biomarkers for neurodegenerative diseases. Specifically, we will take advantage of access to the cerebral spinal fluid (CSF) of patients with normal pressure hydrocephalus, a neurodegenerative disease characterized by progressive cognitive and motor deficits. We will use state-of-the-art proteomics, metabolomics and next generation RNA sequencing technology to identify alterations in the expression of molecular signaling pathways that correlate with clinical signs of the disease. To this end, we will develop supervised and unsupervised machine learning algorithms to extensively mine the gene and metabolite expression data obtained from patient samples. We will specifically test the hypothesis that inflammation is a causative mechanism driving neurodegeneration. Our short-term goal is to identify new CSF biomarkers for normal pressure hydrocephalus, and to establish experimental and analysis pipelines that will allow us to extend our approach to target other neurodegenerative diseases including Alzheimer’s Disease. Our interdisciplinary project is timely and highly innovative, bringing together world-class expertise in neurosurgery, molecular neurobiology, and data science. Preliminary results obtained with OVPR support will allow us to attract long-term financial support through federal grants, private foundations, and industry partnerships.
PI: Alexander Fleischmann, Provost's Associate Professor of Brain Science
Co-PIs: Petra Klinge, Professor of Neurosurgery, Director Pediatric Neurosurgery Division, Director of the Research Center and Clinic for Cerebrospinal Fluid Disorders; Thomas Serre, Associate Professor of Cognitive, Linguistic and Psychological Sciences (CLPS), Director of the Center for Computation and Visualization
Neural Metabolomics and infantile epilepsy associated with mutations in SLC13A5
SLC13A5 deficiency is a newly diagnosed form of genetic epilepsy and developmental delay with seizures beginning within the first days of life. In these patients, homozygous mutations in the SLC13A5 gene, which encodes a plasma membrane citrate transporter result in a severe, early onset multi-focal epilepsy, in addition to cognitive and behavioral symptoms. Progress in finding treatments for this condition has been hampered by the lack of appropriate models to study the brain phenotype. We have already developed a novel model with the human SLC13A5 gene transgene inserted into a mouse with a deletion of the native murine Slc13a5. This mouse expresses only the fully human SLC13A5 in the central nervous system. Furthermore, we have created a knock-in of the human SLC13A5 with pathogenic mutations in Drosophilia. The studies proposed here are for the purpose of creating the full panel of model systems with pathogenic mutations and to obtain preliminary data needed for the submission of a multi-investigator grant to the NIH in order to address key questions: 1) what is the normal function of SLC13A5 in brain physiology 2) how do pathogenic mutations in SLC13A5 result in neural dysfunction and 3) what are the genetic modifiers of SLC13A5 which affect disease expression. This will begin to address our underlying hypothesis that epilepsy associated with SLC13A5 is related to specific neuronal metabolic requirements, which impact neuro-transmitter pools. The identification of the metabolic and neurotransmitter changes may lead to new treatments for epilepsy and also cognitive/ behavioral symptoms associated with SLC13A5.
PI: Judy Liu, Sidney A. Fox and Dorothea Doctors Fox Associate Professor of Ophthalmology and Visual Science, Associate Professor of Neurology, Associate Professor of Molecular Biology, Cell Biology and Biochemistry
Co-PI: Stephen Helfand, Professor of Biology, Vice Chair of Neurology
Dissociating Neurocomputational Mechanisms Underlying Positive and Negative Motivations for Cognitive Effort
When deciding how much effort to invest in a given task, individuals weigh both positive outcomes that could accrue (e.g., praise) as well as negative outcomes such efforts could avoid (e.g., admonishment). However, little is known about the neural and computational mechanisms by which different incentives determine how much and how long we invest effort into cognitively demanding tasks, including whether different substrates exist for different strategies for responding to negative incentives (e.g., working harder vs. more cautiously). Moreover, while both incentives contribute to one's decision to invest effort, a person's relative sensitivity to negative versus positive incentives can significantly impact their health and wellbeing, leading to chronic stress and anxiety. This proposal aims to identify neural substrates underlying cognitive effort allocation in the face of these different incentive types, and measure how differences in sensitivity to these incentives influence a given person's motivation to invest effort in their daily tasks as well as their susceptibility to negative health outcomes. We will leverage a computational model our lab has developed to predict variability in effort investment, combining this model with measures of behavior and neural activity taken while participants perform a novel task. This work will bridge research in neuroscience, economics, psychiatry, and public health, and advance Brown’s position across those fields. Receiving an OVPR Research Seed Fund Award to carry out this foundational research will position us to apply for external funding to make further inroads into what motivates people to exert the effort necessary to achieve their goals.
PI: Amitai Shenhav, Assistant Professor of Cognitive, Linguistic and Psychological Sciences
Co-PI: Debbie Yee, Postdoctoral Research Associate, Cognitive, Linguistic and Psychological Sciences
Smarter AI: Designing Autonomous Systems that Optimize Hardware, Software and Cognitive Components Together
Deep learning (e.g., convolutional neural networks (CNNs)) has gained a lot of attention in the past few years, especially for object identification and classification problems. Despite their strengths, CNNs have several shortcomings, such as their opacity to understand how they make decisions, fragility for generalizing beyond overfit training examples, and inability to recover from bad decisions. These weaknesses play to the strengths of techniques in artificial intelligence (AI) based on generative probabilistic inference, techniques that are inherently explainable, general, and resilient by distribution of many hypotheses representing possible decisions. Unfortunately, probabilistic inference, in contrast to CNNs, is often computationally intractable with complexity that grows exponentially with the number of variables. Combined discriminative-generative algorithms have been proposed as a promising avenue for robust perception by balancing computational complexity with explainability and reasoning, but still may not provide for real-time response, even after careful optimization. Instead, we propose bringing humans back into the loop to provide essential functionality to guide decisions, planning, and management of an AI-enhanced system. In particular, the models that have proven useful for understanding human agency are Causal Bayes Nets. This project seeks to explore how humans and intelligent computers can collaborate effectively by integrating technical functions (i.e., discriminative-generative decision making) with human cognitive processes. A driving theme of this exploration will be complexity (and energy) optimization. Applying more effort and thereby computational and cognitive assistance in a hybrid fashion, and only as needed, will minimize energy consumption, improve response time, and be more likely to optimize users’ needs.
PI: R. Iris Bahar, Professor of Computer Science, Professor of Engineering
Co-PI: Steven Sloman, Professor of Cognitive, Linguistic and Psychological Sciences
Novel analogs of trehalose for treatment of preeclampsia, a devastating pregnancy complication
Trehalose has been found to be an effective treatment for preeclampsia, a potentially life-threatening condition affecting some pregnant women. Unfortunately, administration of trehalose can result in additional complications from undesired bacterial growth and infection from microbes that metabolize trehalose. We propose to prepare analogs of trehalose that retain efficacy in the treatment of preeclampsia but cannot be utilized by bacteria. The prepared analogs will be evaluated using a variety of models for their effectiveness in treating preeclampsia while not serving to promote bacterial growth. The project team of Basu and Sharma provide expertise in carbohydrate synthesis and preeclampsia biology respectively.
PI: Amit Basu, Associate Professor of Chemistry
Co-PI: Surendra Sharma, Professor of Pediatrics (Research), Professor of Pathology and Laboratory Medicine (Research)
Computational Modeling of Blood Flow to Understand Microvascular Dysfunction in Alzheimer's Disease
Alzheimer’s disease (AD), a progressive neurodegenerative disorder affecting millions of people worldwide, and related dementia are becoming the biggest epidemic in medical history. However, AD is a heterogeneous and multifactorial disease, making it challenging to fully understand how the multiple etiologies and age-related prodromal processes in AD contribute to its pathophysiology. Among other factors, deficits in cerebral microvascular structures and functions are recently considered to play a key role in the onset and development of AD. But, it is still unclear whether they are a causal factor for AD pathogenesis or an early consequence of multifactorial conditions that lead to AD at a later stage, despite its importance for early diagnosis and as a therapeutic target. To address this knowledge gap, we performed a longitudinal imaging experiment of tracking progressive microvascular alterations in AD mice for almost their lifespan. Although these data provide us with unprecedently rich information about various cerebral microvascular deficits and cognitive impairment, they were insufficient to determine the cause-effect relationships. Here, in this Seed project we will develop a computational methodology to investigate the mechanistic question. First, we will improve our computational model of microvascular flow and functional hyperemia (Aim 1), and then combine the model with the experimental data for the mechanistic study (Aim 2). The computational model will enable us to essentially “turn on” and “turn off” each microvascular deficit (e.g., thinner vessels, tortuous capillaries, hypoperfusion) and test its effect on oxygen delivery to neurons, which is difficult and sometimes impossible to achieve experimentally.
PI: Jongwhan Lee, Assistant Professor of Engineering
A new direction in building quantum computers: 2D material "lego"
Reducing a material to the atomic 2-dimensional limit has been shown to have profound effects on its properties. Since the successful exfoliation of graphene from bulk graphite , a large family of layered van der Waals materials have been thinned down to a single atomic layer, forming a new material platform covering a wide range of physical properties. The van der Waals assembly technique allows any 2D material to be re-assembled into a designer structure, which has recently led to a flurry of discoveries establishing 2D material heterostructure as a new paradigm for discovering novel quantum phenomena and advancing our understanding of quantum science. Here, we propose a new direction to study an entangled quantum phenomenon called Majorana mode, which is at the heart of topological quantum computation. The PIs plan to develop the capability of thermal transport measurement on materials that are one-atomic layer thin. Measurements of quantized thermal conductance in these materials will offer direct and unambiguous identification of Majorana modes. This effort will build on the established expertise of the PI and co-PI. Prof. Li has developed the necessary techniques of working with 2D materials and performing quantum transport measurements, Prof. Plumb is a leading expert in bulk material growth and neutron scattering, and Prof. Feldman has done pioneering research on Majorana modes in the 2D limit. The proposed effort will establish Brown University as a center for studying and engineering future 2D magnetic material and nano-scale quantum technology.
PI: Jia (Leo) Li, Assistant Professor of Physics
Co-PIs: Kemp Plumb, Assistant Professor of Physics; Dmitri Feldman, Professor of Physics
Search for Topological Waves in Magnetized Gaseous Plasmas
Funds are requested to purchase electronic equipment and cover travel expenses to UCLA's Basic Plasma Science Facility (BaPSF) to make a first observation of a plasma wave of topological origin using the Large Plasma Device (LAPD). The search was inspired by my 2017 Science paper ""Topological Origin of Geophysical Waves"" that reported the surprising discovery that Kelvin and Yanai waves have a topological origin analogous to edge modes in the quantum Hall effect. My collaborators and I have now theoretically predicted, and simulated, other waves of topological origin in a magnetized plasma. We used a realistic model of the LAPD plasma, and the Director of BaPSF (Troy Carter) has granted us time on the LAPD to conduct experiments provided that we can supply the necessary microwave amplifier to drive the waves. If the existence of the waves is confirmed by experiment, it would represent a breakthrough in plasma physics likely leading to both publication in prestigious journals and external funding.
PI: Brad Marston, Professor of Physics
Confronting the Data Deluge using Quantum Machine Learning
One common application of data science across scientific domains is extracting signals from increasingly large data sets. Recent advances in deep learning and artificial intelligence have become a practical necessity for such applications. Given the size of the datasets and the ever growing needs for CPUs for effective training of deep learning algorithms, quantum computing is an attractive solution. Even though there are remaining challenges to make quantum computing devices widely usable, it is important to understand and explore what this new technology and applications of quantum computing algorithms could bring to our fields. We propose to use quantum simulators to perform sensitivity studies for extracting rare signals of new physics in the environment proposed for future particle colliders and free electron lasers. The proposed interdisciplinary work has important implications beyond the disciplines involved as it will develop the technology of quantum machine learning. As quantum computers become available they will allow us to meet computational challenges in many fields and solve scientific questions that are out of reach with current technologies. This seed project will complement and strengthen the core group of faculty members in Physics and Chemistry who explore quantum information science.
PI: Meenakshi Narain, Professor of Physics
Co-PIs: Brenda Rubenstein, Assistant Professor of Chemistry; Peter Weber, Professor of Chemistry
Cosmic History from Mapping the Universe with Neutral Hydrogen
Measuring the intensity of 21 cm emission from neutral hydrogen gas is a novel technique that enables mapping large volumes of the universe in three dimensions. The Tianlai Pathfinder has been carrying out a North Celestial Cap Survey (NCCS). The Tianlai Pathfinder is unique among 21 cm instruments in being able to point and integrate continuously on a limited patch of sky for extended periods. With seed funding Brown could contribute to analyzing this rich data set and produce initial results which will strengthen subsequent proposals. One of the ultimate goals of this research is to measure baryon acoustic oscillations over cosmic time to help understand dark energy, which is currently not understood. This research will also, for example, shed light on the mysterious fast radio bursts (FRBs), detect radio counterparts of gravitational events from sources such as merging neutron stars and lead to a better understanding of galaxy formation. The Tianlai Pathfinder will demonstrate the feasibility of using wide field of view radio interferometers to map the density of neutral hydrogen in the universe after the Epoch of Reionization (EoR). Such radio interferometers are relatively new, and the necessary techniques are still being developed. The Tianlai Pathfinder is also unique in that it consists of two co-located interferometers utilizing different types of antennae (cylinders and dishes), which provides an important opportunity to compare the ultimate performance of these two types of telescopes as the next generation of more sensitive instruments is planned.
PI: Gregory Tucker, Professor of Physics
Is greenspace associated with mental and physical health among pregnant women? A geo-ethnographic exploration
Exposure to greenspace, broadly defined as various forms of vegetation, has been shown to confer various health benefits. Specifically, reducing the likelihood of adverse birth outcomes, reducing the mental health impact of stressful life events, decreasing symptoms associated with behavioral problems, as well as reducing the risk of the onset of obesity. The evidence base of the positive impact of greenspace on health has proliferated to such an extent that in 2015, the United Nations adopted the exposure to greenspace as a sustainable development goal, with target 11.7 stating "By 2030, provide universal access to safe, inclusive and accessible, green and public spaces, particularly for women and children, older persons, and persons with disabilities." With regard to the association between greenspace and birth outcomes, however, the findings are equivocal. The purpose of this multidisciplinary collaboration across public health, medicine, and geography is to attempt to reconcile conflicting findings on the association between greenspace exposure and birth outcomes. By examining the relative contributions of varying measures of green space exposure hypothesized to be associated with mental and physical health during pregnancy, we hope to develop the types of assessments that capture the relevant environmental features that will permit a better understanding of the mechanisms that underpin associations between greenspace exposure and birth outcomes.
PI: Diana Grigsby-Toussaint, Associate Professor of Behavioral and Social Sciences
Co-PIs: Patrick Vivier, Professor of Health Services, Policy and Practice, Professor of Pediatrics, Professor of Emergency Medicine; Kevin Mwenda, Assistant Professor of Population Studies, Associate Director, Spatial Structures in the Social Sciences (S4)
The effect of a driver's license suspension on access to health care
Every year approximately 3.6 million Americans miss or delay health care due to transportation barriers. Though lacking access to a vehicle is the most commonly reported transportation barrier to care, 43 states currently have policies to suspend driver’s licenses as a means of compelling compliance with laws and regulations unrelated to driving (e.g., failure to pay a court fee or appear in court). Approximately 80% of all suspensions are for a non-driving-related offense, and the impact of these suspensions falls primarily on low-income and racial/ethnic minority drivers. Supporters of suspensions view them as one of a limited set of tools for compelling compliance with state regulations. However, little is known about the population of suspended drivers, or the unintended consequences of a suspension for accessing health care. We propose to close this gap using a unique dataset of linked driver’s licensing histories with Medicare and Medicaid claims to provide the first individual level descriptions of this population and how it has changed over time. In the last two years, six states have passed legislation ending non-driving-related license suspensions and several more are considering doing the same. The findings from our proposed work will provide policy makers with the essential information on suspended drivers necessary for developing informed and effective suspension policies and will establish the empirical foundation required for extending our work to estimate the causal effect of a suspension on access to health care, health care utilization and ultimately health outcomes.
PI: Nina Joyce, Assistant Professor of Epidemiology
Co-PI: Andrew Zullo, Assistant Professor of Health Services, Policy and Practice, Assistant Professor of Epidemiology
Co-I: Jasjit Singh Ahluwalia, Professor of Behavioral and Social Sciences, Professor of Medicine.
Improving maternal and child health starting in pregnancy: examining cardio-metabolic risk among women living with and without HIV and their children in South Africa
Each year in sub-Saharan Africa, over 1 million children are born exposed to HIV-infection in utero, but uninfected (HEU). Compared to HIV-unexposed (HU) children, HEU face a range of health consequences including metabolic abnormalities, such as hypertension, dyslipidemia, and impaired glucose function. To date our understanding of factors that affect metabolic health of women living with HIV and their children has been hampered by the vastly different social environments and health status between those living with and without HIV. These differences may lead to important variations in maternal biomedical and psychosocial factors during pregnancy, such as maternal substance use, food security, body mass index, gestational weight gain, which may influence long-term metabolic health for both women and children but have not been explored in HIV-infected populations. This proposal brings together the unique disciplinary perspectives of a psychologist and an epidemiologist to examine how maternal factors during pregnancy influence metabolic outcomes in women and children and to design an intervention to mitigate these factors to be evaluated through subsequent external funding. We propose to leverage the Drakenstein Child Health study, an established cohort of mother-child pairs in South Africa with rich psychosocial and biomedical data. OVPR Seed Funds will be used to examine markers of metabolic function during pregnancy and prospectively evaluate metabolic outcomes in women and children at 5-8 years postpartum. We expect this pilot award to result in a R01 application focused on testing an intervention to improve cardio-metabolic outcomes in women with and without HIV and their children.
Co-PIs: Angela Bengtson, Assistant Professor of Epidemiology; Jennifer Pellowski, Assistant Professor of Behavioral and Social Sciences
Co-I: Stephen McGarvey, Professor of Epidemiology, Director of International Health Institute.