2023 CompX Recipients

About the CompX Faculty Grants Program

Winners of the 2023-2024 Neukom Institute CompX Faculty Grants Program for Dartmouth faculty have been announced for one-year projects. We received over $650K in total requests and awarded a total of $250K.

The program seeks to fund both the development of novel computational techniques as well as the application of computational methods to research across the campus and professional schools.

Dartmouth College faculty including the undergraduate, graduate, and professional schools were eligible to apply for these competitive grants.

Note: * indicates an award that is partnered with assistance from Dartmouth College Research Computing.

+ indicates an award that is partnered with an RA provided through the Scholars program

Art History

Nicola Camerlenghi

THE FINAL PHASE of THE DARTMOUTH VIRTUAL BASILICA PROJECT


Camerlenghi

Camerlenghi

This proposal is the final phase of the multi-year Virtual Basilica Project, which digitally reconstructs one of Rome's most important church—the Basilica of St. Paul Outside the Walls. The aim of this last push is to supplement the existing digital models of the original, fourth-century basilica with a photogrammetric and laser-scanned model of the current building, which replaced the original structure following a fire in 1823. The addition of a scanned model of the modern building will allow users of the desktop and VR versions of the application to time-travel from the present-day to better understand the complexity of the reconstruction and its relation to the original building. Fortuitously, the new building stands on the precise footprint of the old and even their elevations match, which should make for compelling visual transitions from old to new, and vice versa. Indeed, the model of the current building will provide the "entry portal" for the diachronic, virtual experience being offered to students, scholars, and the public, free of charge.

But, aside from its capacity to interact with the existing digital reconstructions, a scanned model of the building would enable users to move at will and investigate the present building in ways not permissible to the average visitor. For instance, one could see distant frescos, up-close; anyone could visit the excavations, which are generally off limits; and, more generally, those who may have difficulty to visit this important site will be able to do so from the comfort of their home computer. Finally, this will be the first time the building is given such treatment. Precise scans offer a priceless baseline for matters of preservation and conservation. To that end, a copy of our scans will be deposited at the Vatican.The scans themselves will be conducted with drones and fixed stations in collaboration with surveyors at Roma Tre University during July 2023, in conjunction with events related to the 200th anniversary of the tragic fire

Biology & Computer Science

Kathryn Cottingham & James Mahoney

Exploring Watershed Moments Using Virtual Reality


Cottingham

Cottingham

Kathy Cottingham, James Mahoney, Quin Shingai, Monika Roznere, Jess Trout-Haney, Jeff Kerby, John Bell and the DEV studio

Freshwater resources sustain life while serving as economic and cultural infrastructure for societies worldwide, providing food, transportation, drinking water, recreational opportunities, and other valuable ecosystem services that promote human well-being. Despite the critical importance of water to civilization, humans often adversely impact water quality via deforestation, fertilizer overuse, and pollution - impacts that can be exacerbated by the indirect effects of climate change. Unfortunately, the combined direct and indirect anthropogenic effects on lakes and reservoirs can be difficult to understand due to the separation of cause and effect in space and time. We seek to raise awareness of these issues by harnessing virtual reality (VR), a tool which can enhance users' empathy and modify their subsequent behavior.


Mahoney

Mahoney

In this study, we aim to develop a working prototype for an open-source VR simulation of a generalized freshwater lake ("Sim Lake") to serve as a base environment that can be adapted for research, education, and outreach. Specifically, we plan a two-stage, coupled prototype that combines VR simulations of the area around a lake with stereoscopic, photo-based components, above and below the water surface. In stage 1, we will simulate the experience of visualizing abstracted watersheds from a bird's eye view, and also standing on a lake shore, creating an improved version of similar work in Dartmouth extended reality courses. In stage 2, the simulated environment will provide context for a more photorealistic world populated by stereoscopic images closely informed by real-world above water and underwater photos and videos. This more realistic view will facilitate the examination of in-lake conditions, and would include the ability to "tune" properties that cue participants as to the potential water quality and safety of the simulated freshwater lake ecosystem. For example, we can change the clarity and color of the water from a clear blue signaling "clean" water, to a murky blue-green signaling a cyanobacteria bloom, to a murky brown signaling transport of sediment into the lake following a large rainfall event. We can alter the lake bottom from cobble to sand to mud or add and remove particular organisms such as lake trout or invasive Eurasian watermilfoil. Changes to the surrounding land can also be connected to what things look like underwater. Upon completing beta-testing and debugging, the environment will be available open source via a project website, for use by our team and others. Once developed, we plan to gamify "Sim Lake" to communicate the importance of freshwater resources and help users to better understand how their activities on land impact the freshwater ponds, lakes, and reservoirs on which they depend. We also hope to use this tool in research to improve human use and management of these resources. 

 

Linguistics

Christiane Donahue *

"Medium" Data Work: Studying Student Writing Growth via Portfolios


Taine Donahue

Taine Donahue

In collaboration with Research Computing, this CompX grant will support the quarterly draw of digital portfolio data in the DartWrite research project for the 2023-2024 academic year. This project will pursue, with data already collected, a specific project or more that demonstrate the interest and value of the project's data, and will share results to identify specific fruitful research questions for current and future projects.

Psychological and Brain Sciences

Brad Duchaine

DNN modeling of category-selective areas in developmental prosopagnosia 

Brad Duchaine

Brad Duchaine

 In this project, we aim to accelerate research into Developmental Prosopagnosia (DP), a neurodevelopmental disorder causing severe face recognition deficits. We propose to use Deep Convolutional Neural Networks (DNNs) to create computational models of category-selective brain regions. These regions respond selectively to specific visual categories such as faces, places, bodies, and words. 

Our central hypothesis is that by combining intensive scanning of individual participants, high-resolution analysis of responses to individual images, and DNN-based computational modeling of neural responses, we can greatly speed up DP research. We plan to test the accuracy of these models in predicting responses in four individuals with DP. If successful, this approach could transform our approach to investigating visual recognition deficits and could be extended to other neurodevelopmental and psychiatric disorders. 

Our project has two primary objectives: 

  1. We aim to determine if models of DP category-selective areas can accurately predict responses to images not used in model training. We will compare the predicted and observed correlations with noise ceilings, which represent the theoretically maximum expected correlation. 

  1. We will assess the ability of DP models to predict responses to images designed to test theoretically-motivated questions about DP. We will present stimuli addressing two key theoretical questions to the DPs and compare the predicted and observed responses. 

If our results are promising, they will serve as preliminary data for an R01 proposal. This proposal will test a larger sample of DPs (n=30) to convincingly demonstrate the value of our approach. 

Psychological and Brain Sciences

Lucas Dwiel

Stimulating brain networks to direct psychedelic induced plasticity


Dwiel

Dwiel

Recent clinical studies have reported beneficial outcomes from pairing psychedelic drugs with psychotherapy. This project will test the theory that the biological mechanisms underpinning these encouraging results will, similarly, create longer-lasting changes in brain activity and behavior when such drugs are paired with brain stimulation. To begin probing the physiological and behavioral consequences of pairing brain stimulation with psychedelic drugs, while avoiding the difficulties associated with human psychedelic trials (e.g., the challenge of an adequate placebo control and strong expectation bias), we have completed preliminary studies administering lysergic acid diethylamide (LSD) to rats. We have shown that LSD alone has acute, detectable effects on brain activity compared to saline; yet, during the known window of enhanced plasticity (~24 hours later) there were no differences in frontal-striatal local field potential oscillations compared to animals given saline, as determined using a machine-learning approach. However, when the rats' infralimbic cortex (IL) was stimulated, this induced more changes in brain activity in the rats that were given LSD 24 hours before stimulation, compared to those given saline 24 hours before stimulation. This physiological evidence supports our hypothesis that psychedelic drugs enhance the ability of brain stimulation to alter frontal-striatal activity in awake, freely-moving rats.

This CompX grant will allow us to implement the Neuropixels system in order to characterize the interaction of LSD and brain stimulation on corticostriatal brain states in rats during a clinically relevant impulsive decision making task. The Neuropixels system will provide high-dimensional single-neuron spiking data and local field potentials (LFPs) from multiple brain regions implicated in decision making. These data will allow us to use computational methods to characterize the effects of brain stimulation and LSD on brain states and their dynamics.

Chemistry

Chenfeng Fe

Accelerated discovery and fabrication of robust 3D-printable slide-ring hydrogels


Fe

Fe

Natural materials are endowed with exceptional mechanical properties to survive under various extreme conditions. These biomacromolecules and their assemblies are arranged hierarchically to form homo- and heterogeneous 3D architectures across the nano to macro-scales. While defects are ubiquitous in biological systems, biomaterials are more tolerant to defects because of their capability to rearrange in response to stress. In comparison, parts manufactured using synthetic polymers often possess shorter life spans due to the intrinsic nature of polymer networks, material inhomogeneity, as well as defects generated during the manufacturing process. The lack of a comprehensive structure-property-process relationship at the fundamental level limits the advancement of soft materials. Hence, there is an emergent need to weld precise control of soft materials' chemical structures, functions, and hierarchical placements.

To address this problem, we plan to develop a systematic approach by combining the bottom-up slide-ring polymer design with the top-down direct ink writing (DIW) to develop robust polymer hydrogels that are tough and fatigue-resistant. Our work capitalizes on connecting the bottom-up, slide-ring polymer design with the top-down detection and correction of in-line defects during additive manufacturing (AM) through high-throughput synthesis and machine learning (ML) approaches. The research outcome will enable the development and fabrication of next-generation defect-free slide-ring-based resilient materials.

Anthropology

Brinker Ferguson

Low Cost Rapid Digital Preservation Documentation Workflows for At-risk Historic Wooden Architecture in Puerto Rico


Brinker Fergason

Brinker Fergason

In recent years Puerto Rico has responded to massive destruction caused by natural disasters with an equally powerful conservation response. In 2022, the Puerto Rico State Historic Preservation Office identified over 100 culturally important wooden vernacular homes as some of the most vulnerable and threatened built heritage on the island. These 19th and 20th century historic properties are unique combinations of European and Creole-vernacular architecture which include the adoption of architectural elements and motifs from other sister islands in the Caribbean. A state-wide "Wooden Architecture Restoration Initiative"  is slated to begin in 2025, but before restoration projects can officially start, precise architectural drawings are needed for baseline conservation documentation. Currently, conservation documentation in Puerto Rico focuses on traditional hand measurements that is costly in both time and expertise. This will delay project start dates making these sites even more vulnerable to incoming hurricanes, which are becoming increasingly severe due to climate change. Thus, developing an accurate and rapid conservation documentation workflow with low-cost and accessible technologies for the creation of architectural drawings is critical for the earlier protection and preservation of these important Puerto Rician cultural sites.

To meet these needs, Dartmouth College in partnership with Centro de Conservación y Restauración de Puerto Rico (CENCOR) and the School of Architecture at the Polytechnic University of Puerto Rico, have been asked to develop and test three new conservation preservation workflows using the latest in digital capture technologies to see how they can increase speed and accessibility compared to traditional hand measurements. The goal of this research is to make headway towards the democratization of digital capture research within the field of cultural heritage preservation.

Classics

Julie Hruby

Associating Fingerprint Patterns with Age and Sex: A Quantifiable Approach


Julie Hruby

Hruby

A wide range of prehistoric and ancient Greek ceramic objects, including vessels, ceramic sculpture, seal impressions, and writing tablets preserve the fingerprint impressions of their producers. Traditionally, archaeologists have matched prints in an effort to understand ancient labor systems, but more recently, we have also begun to ask a much wider range of questions. The ages and sexes of producers are among those questions, but so far, the techniques that have been used to reconstruct those factors have typically been able to work on the level of populations rather than individuals, and they have also been subject to challenges posed by differential clay shrinkage rates.

The current project will improve the accuracy of sexing and aging producers of ancient Greek artifacts by using fingerprints that were accidentally impressed in objects made by modern Greek ceramicists as a reference sample. Fingerprint impressions from modern Greek adult potters of known sexes and age grades have already been collected and scanned with a high-resolution 3D scanner, and a Greek attorney has assisted us in complying with both European Union and Greek law as they relate to the collection of prints from juveniles. We have also begun the process of collecting and scanning modern prints from juveniles, and we will begin archiving our raw data.

Chinese Language and Literature

Hua-yuan Li Mowry

Leveraging Digital Platforms to Compile Multimedia Chinese Language and Culture Materials in the HTML Format


Mowry

Mowry

The animation film The Buffalo Boy and His Flute (released in 1963, 22 minutes) is uniquely Chinese: It is in the traditional Chinese water-ink painting style, and has been hailed as a miracle of animation. The Buffalo Boy and His Flute is intentionally delivered as an "international" animated film (i.e., without the limitation of dialogue or exposition in any language, so viewers can interpret and appreciate the film as a primarily visual product) suited to language teaching/learning purposes. I have completed a good portion of the required archival research on, and in-person interviews about, the film's production background—cultural, socio-political, cinematic, artistic, and esthetic. With the support of a 2018-19 CompX grant, my collaborators and I also accomplished about two thirds of the work that is computational in nature. This includes completing the "read along" feature, embedded vocabularies, voiceover and subtitle annotation. Further computational tasks yet to be performed include expounding upon grammatical points visually with clippings from the film, integrating the project's various components into an artistically attractive and pedagogically sound whole, and associated final-phase miscellany.

In addition to The Buffalo Boy and His Flute, we plan to convert a Captivate (v.8) generated digital textbook for intermediate to advanced Chinese language classes, the Snapshots of China, into HTML format. The traditional "paper" version of Snapshots of China, in 30 lessons, has been used as one of the main textbooks at Dartmouth and Dartmouth's Beijing FSP/LSA+ programs since it was written/compiled in 2013. In 2016-17, with the help of two able CS Ph.D. students, we authored the digital version. The creation of the courseware was timely during the first two years of the Covid-19 pandemic, when almost all Dartmouth classes were taught online. While we are not debating the pros and cons of virtual/online classes, my collaborators and I firmly believe that digitalization—and instructional and educational technologies—have positively impacted both the field of education and that of publication.

Earth Sciences and Thayer School of Engineering

Sarah Slotznick and Geoffroy Hautier

From Earth to Mars: Atomic Level Probing of Hydrohematite's Magnetism


hautier

hautier

Iron minerals are a critical record of planetary processes from planetary formation to plate tectonics to surface processes like nutrient cycling and environmental conditions in both modern and ancient times. In addition to being the most abundant transition metal in terrestrial planets and their crusts, iron has two fundamental properties: a) it is redox sensitive, cycling between +2 and +3 valence states and b) it is ferromagnetic. These properties translate to iron minerals which are being utilized to understand the history of environmental change on Earth as we consider future anthropogenic climate change. Extension to other terrestrial planets like Mars is the next frontier. From a technological perspective, magnetic materials are vital for the alternative energy transition and investigating natural systems could spark new ideas for material synthesis.


slotznick

slotznick

This project will focus on understanding the magnetism of hydrohematite – a version of the common iron oxide hematite that contains structural water. Hydrohematite could potentially be an unrecognized water reservoir in arid planetary surfaces like Mars and detailed study could provide new clues into both natural magnets and the planet's evolution. The team will perform magnetic measurements and synchrotron-based spectroscopic analyses on natural and synthesized specimens. These results will be compared with computational methods at the atomic and nanometer scale using first-principles calculations and micromagnetic modeling to identify unique magnetic properties of hydrohematite and reveal the physics behind why these occur.

Psychological and Brain Sciences

Adam Steel

Revealing the Function of the Human Anterior Temporal Lobe with Machine Learning


Steel

Steel

The human anterior ventral temporal lobe plays a critical role in memory formation. Prior work has shown that this area, situated on brain's bottom surface, harbors our memories of our family member's faces, sentimental and important objects, and places dear to our hearts. Critically, the anterior ventral temporal cortex is among one of the regions most profoundly impacted by Alzheimer's Disease and frontotemporal dementia, so understanding the functional composition of this area will be necessary to develop therapeutics and interventions for these devastating illnesses. However, our understanding of this region's computational role is limited by technical challenges. Specifically, the most widely used brain imaging technique, functional magnetic resonance imaging (fMRI), fails to adequately image this area because of signal dropout caused by magnetic field distortions. This severely impedes neuroscientific investigation of this brain area and human memory more broadly.

 

In this project, our central aim is to optimize fMRI acquisition and data processing to study the anterior ventral temporal lobe. First, we will explore novel, advanced multi-echo fMRI acquisition to improve signal dropout artifacts in this region. Second, we will develop new signal processing algorithms to further improve data quality and model fitting. Finally, we will apply this optimized protocol combined with new experimental paradigms to study the computational principles underpinning anterior ventral temporal lobe function. Overall, this project will pave the way to further study of the anterior ventral temporal lobe and reveal new insights into human memory in health and disease.

Computer Science and Geisel School of Medicine

Soroush Vosoughi and Xiaofeng Wang

Epigenomic–Transcriptomic Modeling with Multi-Omic Data and DNA Sequences Leveraging Graph Neural Networks


soroush-vosoughi

soroush-vosoughi

Understanding enhancer-target interactions (ETIs) is critical for unraveling the complexities of the human genome. Current experimental methods have limitations, and machine learning (ML) models often overlook the dynamic nature of ETI networks. Our research proposal aims to address this issue by leveraging graph neural networks (GNNs) to predict ETIs, using multi-omic data and DNA sequences.


Wang

Wang

 

Our approach enables the examination of cRE regulation of gene expression and the revelation of how epigenomic profiles of cREs contribute to DNA interactions and gene expressions. This could lead to reduced time and resource consumption for hypothesis construction and promote research in cancer epigenome studies and disease prediction applications. To achieve our objective, we first construct a GNN-based cRE-interaction prediction model with multi-omic data and DNA sequences. Next, we construct an ETI prediction model to jointly examine cRE interactions and gene regulations. Finally, we construct a hypothesis-establishment and validation framework based on the ETI prediction model. By accomplishing these tasks, we expect to gain a better understanding of how intergenic cREs regulate target genes and establish high-quality hypotheses that biological experiments can verify.