2020 Projects

Anthropology

Megan Reyes ’21, Caroline Klureza ’22 and Professor Zachary D. Cofran (Anthropology)

Symposium Poster and Presentation Audio

Project Abstract

The human pelvis is uniquely adapted for bipedal walking, and the shape of the ilium plays an important role in locomotion. Shape changes to the human ilium during individual growth and development reflect biomechanical adaptations. Australopithecus africanus from South Africa walked bipedally around 3 million years ago, but their ilia have some shape attributes similar to both humans and non-human apes. It has previously been proposed that A. africanus ilia followed a growth trajectory more similar to apes than to humans. However, that research was based on methods that do not provide high-resolution shape information. We revisited the issue of ilium growth using advanced geometric morphometric methods, to test the hypothesis that A. africanus ilia follow the same shape growth pattern as humans.

We established the human growth trajectory using 40 individuals in various stages of development from infancy to adulthood. We quantified ilium shape with 148 three-dimensional landmark coordinates for each individual. We obtained a 3D growth trajectory with a principal component analysis, characterizing shape differences among human ilia that were caused by growth. Using this human growth projection, we “grew” two immature A. africanus ilium fossils into simulated adults, which we then compared to an actual adult A. africanus fossil. To test our hypothesis, we also simulated development of immature human ilia and compared the simulated versus actual adult shape.

Results were conflicting: one simulated A. africanus adult suggested the species may follow a human growth pattern, whereas the other simulated adult suggested a different pattern. This discrepancy could be due to many factors. On the one hand, A. africanus ilia may have grown differently from humans, as previously suggested. Alternatively, the limited fossil sample may not be ideal representatives of the species, perhaps due to distortion or incorrect reconstruction.

Biology

Symposium Poster and Presentation Audio

Project Abstract

Translation initiation is the process that assembles the ribosome, the molecular apparatus which translates the genetic code and synthesizes the corresponding protein. Assembly of the ribosome on a specific mRNA during initiation is an important step for regulating translation. Initiation begins with the assembly of a pre-initiation complex (PIC) in which the small ribosomal subunit is joined by several protein initiation factors (eIFs). The PIC attaches to the 5’ end of an mRNA strand and scans for the start codon. Once the start codon is reached and identified, scanning stops, and the full ribosome is assembled. The largest and most complex of the eIFs is eIF3, which participates in each phase of translation initiation yet remains poorly understood. In the yeast Saccharomyces cerevisiae, eIF3 is made up of 5 subunits, identified as eIF3a, b, c, g, and i. To investigate the role of eIF3 in translation initiation, we employ ribosome profiling, which enables us to learn the position of each translating ribosome on every mRNA in living cells. We use ribosome profiling to monitor the effects of specific mutations to the eIF3 complex and investigate the features of the specific mRNAs most sensitive to these mutations. We are focusing on two mutations to eIF3: one disrupting the association of eIF3i and eIF3g with the rest of the eIF3 complex (DDKK) and another that disrupts the entire eIF3 complex (Degron). By comparing the mRNAs sensitive to both DDKK and Degron mutations to those mRNAs uniquely sensitive to the Degron mutation, we hope to disentangle the contributions of eIF3i and eIF3g from those of eIF3a, eIF3b, and eIF3c. This analysis reveals that mRNAs that appear more sensitive to disruption of eIF3i and eIF3g have longer 5’ UTR lengths and are more sensitive to mutations of the helicases Ded1 and eIF4A, or to the initiation factor eIF4B. These observations are consistent with a growing body of evidence that shows that subunits eIF3i and eIF3g play important roles in the scanning step of translation initiation.

Symposium Poster and Presentation Audio

Project Abstract

Translation initiation is a complex pathway in eukaryotes and is the step by which translation is primarily regulated. The misregulation of translation initiation is a hallmark of cancer and other human diseases. Translation initiation requires the formation of the pre-initiation complex (PIC) which then attaches to the 5’ end of messenger RNA (mRNA) and then scans the mRNA in order to locate the start (AUG) codon. Translation initiation involves at least twelve different eukaryotic initiation factors. Eukaryotic initiation factor 3 (eIF3) is the largest, most complex, and least understood of these initiation factors. eIF3 participates in each phase of translation initiation and plays a large role in mRNA attachment, scanning, and start-codon recognition. In Saccharomyces cerevisiae, eIF3 is made up of 5 subunits: eIF3a, eIF3b, eIF3c, eIF3g, and eIF3i. All five of these eIF3 subunits are overexpressed in different cancer cells and are causally linked to cancer development. We combined two techniques to monitor translation in living cells. One is called ribosome profiling, which allows us to locate the position of every translating ribosome on every mRNA molecule in living cells. The other is called mRNA sequencing which is used to thoroughly determine the abundance of each individual mRNA. By comparing the information from these two approaches, we can calculate the translational efficiency of individual mRNAs and how these are impacted by mutations to eIF3. We then investigate the features of mRNAs most sensitive to specific mutations of eIF3 to infer the mechanistic contributions of eIF3 during translation initiation. I am focusing on how the presence of upstream open reading frames (uORFs) on mRNAs impact their sensitivity to mutations to eIF3. A uORF is a translatable stretch of RNA defined by a start and stop codon that is located upstream of the main coding sequence of an mRNA. uORFs force the PIC, as it is scanning, to make decisions as to whether or not it has found a start codon. Cognate uORFS begin with the universal AUG start codon, whereas near-cognate uORFs begin with a start codon that differs from AUG by one nucleotide (e.g. CUG or UUG). We observe that a mutation that disrupts the interaction of eIF3i and eIF3g with the remainder of the complex (DDKK) and a mutation that disrupts the entire eIF3 complex (eIF3 degron) both result in increased translation of near-cognate uORFs as compared to cognate uORFs. This effect is most severe on mRNAs that are sensitive to both mutations. We are currently testing these effects with statistical models. Taken together, these observations are consistent with a model in which eIF3 contributes to discrimination against near-cognate start codons, with the eIF3i and eIF3g subunits playing an important role in this discrimination

Symposium Poster and Presentation Audio

Project Abstract

The global importance of microbes as beneficial and pathogenic organisms cannot be overemphasized. A variety of modern tools are available for characterizing communities of microorganisms and their genomes through genomics and metagenomics. In genomics, whole genome sequencing, genome assembly, and downstream analysis can be used to characterize pathogenicity, while the metagenomics QIIME2 pipeline can be used to characterize entire microbial communities. The aim of the current study is to learn and apply the practical assembly of bacterial genomes and the QIIME2 pipeline. Whole genome assembly and analysis were learned and applied through a case study of 18 Salmonella strains isolated from 138 fecal samples of Sokoto Gudali cattle at the University of Ibadan in Ibadan, Nigeria. The DNA was sequenced using Illumina MiSeq, assembled de novo using FastQC, Trimmomatic, SPAdes and Quast, and subjected to downstream analysis by Pathogenwatch, MLST, ABRicate, CGE, FigTree, Microreact, PATRIC, and BAGEL4. The metagenomics pipeline QIIME2 (v. 2020.6) was studied through the online “Moving Pictures” tutorial. The 18 Salmonella strains were of the species Salmonella enterica ssp. enterica belonging to five different sequence types ST5332, ST473, ST519, ST5330 and ST3961 and serovars Banalia/Tounouma, Hadar, Koketime, Hermannswerder, and Chomedey/Glostrup, respectively. There was an absence of virulence genes in all strains and 83.3% of the strains carried only one antibiotic-resistant gene, aac(6’)-laa, which suggests that these Salmonella strains pose minimal health risks. Genome assembly and analysis, as well as the QIIME2 pipeline, were learned successfully based on the same results obtained from two different operating systems. These two techniques provide insight into bacterial genomes and communities that proves valuable in an age of increasing concern about infection and disease.

Symposium Poster and Presentation Audio

Project Abstract

Cnidarians, such as corals and anemones, reproduce through highly synchronized mass spawning events induced by temperature and lunar cues, but little is known about the biological mechanisms involved in this synchronization. We are interested in whether gonadotropin hormones play a role in controlling the timing of spawning, as they do in many other animals, including humans. We used the model organism Nematostella vectensis, a sea anemone, to examine the baseline timing of spawning, and to begin to determine if these reproductive hormones have an influence on spawning. To establish the baseline timing of spawning events we conducted 5 trials, each with more than 20 anemones, during which we exposed the anemones to light and temperature cues and collected photo data of the spawning event over about a 15 hour period for each trial. We found that most spawning happens between 8.25 and 11 hours after the cue, with an average of 9.42 hours. We conducted one additional trial investigating exposure to a gonadotropin hormone and found that anemones treated with hormones spawned more frequently than anemones in a control treatment. These data will be used to determine the effect of reproductive hormone signaling in the synchronization of cnidarian spawning.

Symposium Poster and Presentation Audio

Project Abstract

The embryos of little skates face a daunting task: develop in their egg case at the bottom of the ocean for up to a year. As they develop, they must exchange respiratory gases with the world outside, and they do so by actively ventilating their egg with a specialized embryonic appendage at the tip of their tail that pumps water in through any one of the egg capsule’s four tubular tendrils. This mechanical work takes energy, and the only nutrition that they have while in the egg is their yolk, which must also power metabolism and growth. Thus, ventilation presents the embryo with a mission-critical cost-benefit trade-off between the energy required and the functional benefit of the behavior. What are behavioral algorithms that the embryos employ to ventilate successfully? In embryos at different stages of development, we observed changes in their behavior-environment relationships that may correlate with the gradual expansion of their Umwelt, the subjective sensory information that they integrate from their environment. One challenge is that the egg capsule is initially closed to the environment: no ventilation is possible. Each tendril has a small slit that is blocked by a jelly-like substance; only after the jelly is dissolved by enzymes from the embryo can that tendril be used for ventilation. In the intermediate case, where some but not all of the tendrils are open, the embryo must decide which to use. Pumping in a closed tendril expends energy without a functional benefit of gas exchange. How might an embryo, with an incomplete Umwelt, decide? We propose a simple stochastic model. Specifically, the on-going decision of which tendril to use is randomized. We show how this stochastic model may allow for sufficient long-term gas exchange with little extra cost compared to the deterministic model that assumes a complete Umwelt and perfect knowledge of the system.

Chemistry

Symposium Poster and Presentation Audio

Project Abstract

Transition metal dichalcogenides (TMDs) have recently become the focus of increased attention for their use as two-dimensional materials in digital electronic and spintronic applications. Their properties can potentially be modified and optimized for specific applications via doping or applying strain. The purpose of this study is to observe the effects of substitutional Mn-doping and strain on the magnetic, optical, and electronic properties of a monolayer of MoSe₂. Specifically, this investigation monitored the sensitivity of the band gap, the magnitude of the local magnetic moments, and the driving force for ferromagnetic ordering. Compressive and tensile strains of up to 15% were applied, and the monolayers were doped at concentrations of 6.25% and 12.5%. This study was done with density functional theory (DFT) calculations using the PBE and HSE exchange-correlation functionals. We found that the band gap tended to decrease in magnitude with increasing percentages of both compressive and tensile strain; this pattern was observed with and without Mn-doping. Pure MoSe2 is a direct-gap semiconductor, but PBE predicted that most doped structures would behave as a half-metal, with a transition to fully metallic with increased compressive and tensile strain. HSE predictions retained semiconductor character for many of the structures, with a transition to metallic character with increased compressive and tensile strain. Additionally, we found that the local magnetic moments on the Mn dopants increased with increasing tensile strain or decreasing compressive strain. There is also a consistent energetic driving force for ferromagnetic ordering relative to antiferromagnetic ordering, with an increased energetic difference at increased tensile strain (and decreased compressive strain). Our study demonstrates the potential for engineering Mn-doped MoSe2 semiconductors with tailored magnetic and electronic properties.

Symposium Poster and Presentation Audio

Project Abstract

In recent years the study of the electronic and luminescent properties of trinuclear gold(I) complexes (CTCs) has led to applications in the field of organic light emitting diodes as well as thin film development for anti-corrosion materials. However, possible usage of these CTCs as catalysts has not been fully explored. A well known reaction type for gold(I) centers is oxidative addition, which takes a gold(I) center and converts it to a gold(II) or gold(III) center, depending on the number of electrons transferred. There have been previous experimental studies which have demonstrated that CTCs are capable of undergoing oxidative addition with dihalogens through a stepwise process, leading to mixed valent gold(I)/gold(III) centers. In addition to these experimental studies, there have also been reports of crystal structures with the addition of methyl iodide to gold centers, but there have been no reports or studies on the mechanistic details of these reactions. In this study, we aim to explore the kinetic and thermochemical properties of the oxidative addition of methyl iodide to a CTC. The initial steps of this work focused on determining the best combination of functional and basis set to model the bond length and vibrational frequency of the carbon-iodine bond in methyl iodide. This combination was then used to determine the mechanism of action for the oxidative addition of methyl iodide to the CTC used in the study. Through this study, it was determined that a hybrid density functional and a fully augmented basis set with a pseudopotential on iodine was needed to accurately model the carbon-iodine bond (B3PW91 and aug-cc-pVTZ). Due to the solid state structure of the CTCs, dispersion corrections were added for the modeling of the species involving the CTCs. The energetic and structural results are presented for the benchmarking and mechanistic studies performed for this study.

Sonia Santos ’21 Linnea Martin ’21 and Professors Sarjit Kaur and Edith Stout (Chemistry)

Symposium Poster and Presentation Audio

Project Abstract

Ambers are produced over millions of years of resin fossilization and are composed of a mixture of terpenoid compounds. Terpenes, a general category for organic compounds found in nature, can be broken into subcategories of mono-, sesqui-, di-, and triterpenes. Terpene composition of ambers can be determined through the analytical technique gas chromatography-mass spectrometry (GC-MS). Infrared spectroscopy (IR) is also helpful in comparing spectra to an in-house database of IR spectra of ambers of known geological origin studied previously. Through chemical analysis of amber, insight is gained regarding the botanical source as well as the geographic distribution and evolution of tree species. In this study, these two analytical techniques, IR and GC-MS were used to analyze, characterize, and compare terpenes found in four geological Arkansas amber samples. Three of the four samples had similar IR scans, showing characteristic peaks for ambers and are closely related. The other amber sample, while showing some organic fossil-like peaks, very likely contained a silicate contamination and precluded a comparison to the other ambers. Based on GC-MS, mono-, sesqui-, and triterpenes were present and, therefore, all four samples are designated triterpene ambers. It is unusual to find triterpene ambers in the Arkansas region as most triterpenes (C30) found in ambers come from trees in the dipterocarpaceae species which are prevalent in Southeast Asia and India. On the average, all four samples are similar based on the chemical composition of sesquiterpenes and triterpenes, after allowing for some difference expected due to extraction of organic components and contamination of the one amber. Further evaluation of the GCMS data is in process for a more detailed comparison of the major components in each amber sample.

Symposium Poster and Presentation Audio

Project Abstract

Propolis, a resinous substance produced by honey bees from various plant resins and other botanic substances, is used by bees as a structural and disinfectant component of the hives. Its antioxidant, antimicrobial, antifungal, antiviral, and anti-inflammatory properties make it a subject of interest for its potential applications in human medicine and health. For this project, the chemical components of propolis collected from a local beehive were determined by a variety of methods, including sublimation, GC-MS, HPLC, and NMR. Numerous compounds were isolated and identified within the propolis samples, including several known antioxidants, with composition variation between samples. Antioxidant assays such as DPPH radical scavenging were used to determine the antioxidant potential of the propolis. The propolis was found to have notable radical scavenging activity, several times more effective than the reference antioxidant, butylated hydroxytoluene (BHT).

Symposium Poster and Presentation Audio

Project Abstract

Fullerenes are a novel allotrope of carbon that consist of empty, spherical carbon cages of varying size. Metallic nitride fullerenes (MNFs), M3N@C2n, are an especially stable family of endohedral metallofullerenes (EMFs) that have been shown to have unique electronic and chemical properties relative to empty fullerenes and other EMFs. Functionalizing the exohedral surfaces of MNFs has been shown to enhance the physical properties of the compounds, but successful functionalization is relatively limited due to the variance in electronic properties exhibited by different encapsulated metal species and different carbon cage sizes. In this research, MNFs of interest (M=Sc,Gd,Lu) were synthesized from their corresponding metal oxide, M2O3, using an arc vapor deposition reactor under a He and N2 atmosphere. In the reactor, high current was run between an interchangeable metal oxide packed graphite anode and a simple graphite cathode to produce a hot plasma that vaporized the packing material and subsequently deposited a carbonaceous soot consisting of both empty fullerenes and MNFs on the reactor lid interior. Upcoming work on this project will entail the use of the Bingel, Prato, and Diels-Alder reactions to potentially produce novel functionalized MNFs, once significant quantities of desirable MNFs have been produced and isolated.

Computer Science

Symposium Poster and Presentation Audio

Project Abstract

This project tracked temporal and spatial variation in sentiment within tens of thousands of books. This research aims to highlight relationships between historical events and locations, explored through sentiment expressed within prevailing literature. To understand the emotions elicited from each book, we used two lexicons to process the texts, both of which produced intensity of emotions ranging from 0 to 1. NRC-EIL was used to gather data on eight categories of emotion, while VADER produced a single sentiment score that ranged from negative to positive. The metadata available for each book – based on previous URSI research – provided the year the book was published and locations associated with the author, according to Wikidata. With all our gathered data, we calculated average sentiment scores and mapped our data in different ways according to the information we had available: the locations mapped were cities, US counties, US states, and countries. These maps were also divided so that each location choice would show sentiments in total, with interactive buttons to show each of the eight emotions as well as positivity and negativity. An optional time slider enabled the map to show data according to a chosen year. These visualizations showed somewhat stable sentiments, but interesting changes were seen during turbulent decades. We expanded our investigation into sentiment by calculating outliers among positive-to-negative sentiments for countries and world cities. Studied per year, these outliers were graphed as histograms or as scatterplots. We calculated outliers for all of our data as well as a stricter sample size. These visualizations also showed similar sentiment values as there were not many outliers. We may continue to analyze literature in different languages to explore the differences in sentiment among language groups.

Merrick Chang ’21, Furrukh Asif ’22, Sudais Moorad ’21, and Professor Luke Hunsberger (Computer Science)

Symposium Poster and Presentation Audio

Project Abstract

The purpose of this URSI project was to help establish a (virtual) Temporal Reasoning Laboratory at Vassar College. The principal goal was to create a public repository that will include implementations of recently developed algorithms for manipulating Simple Temporal Networks (STNs) and Simple Temporal Networks with Uncertainty (STNUs). The algorithms were drawn from the recent literature on Temporal Networks. The public repository will enable researchers in the field to use the implemented algorithms to carry out reproducible empirical evaluations of the algorithms.

Temporal Networks are data structures for reasoning about temporal constraints on events in a variety of circumstances. For example, a computer agent responsible for planning a set of activities might use a Simple Temporal Network to represent the starting and ending times of those activities, as well as constraints on their durations, deadlines, and inter-activity constraints. A consistent STN is one that has a solution (i.e., a fixed schedule for doing those activities that is guaranteed to satisfy all of the constraints). Efficient algorithms are available for the computer agent to check the consistency of the STN, and for generating a variety of solutions. More complex problems involve scheduling activities in real-time (i.e., as opposed to fixing a complete schedule in advance). Algorithms for converting a consistent STN into dispatchable form ensure that real-time execution can be guaranteed to succeed with minimal computational overhead.

An STNU augments an STN to include contingent links that can be used to represent actions with uncertain duration. For example, our computer agent might control when it gets into a taxi, but not the duration of the taxi ride to the train station. Determining whether the activities in an STNU can be successfully executed in real-time is a more complex problem. However, assuming that the uncertainty is bounded by known values (e.g., the taxi ride might take between 10 and 25 minutes), efficient algorithms can check whether an STNU is dynamically controllable. Such algorithms allow an agent to use a dynamic strategy for deciding when to start actions, meaning that the agent’s decisions can react to observations of the durations of contingent links.

Nick Weiner ’22, Jemma Brooker ’23, Ryan Hornby ’21, and Professor Jason Waterman (Computer Science)

Symposium Poster and Presentation Audio

Project Abstract

Smartphones use privately generated data to make the applications we use more convenient. For example, Spotify uses listening history to generate suggestions for what to listen to next. Applications can also collect location, sensor, call history, and biometric data to give more personalized results. This convenience comes with a price - some of these applications collect data in unexpected ways, such as to create a profile for their users, to sell data, or to use it for targeted advertisements. Aegis is a proposed solution to prevent illicit use of private smartphone data. Aegis is a plugin for Flutter, a cross platform toolkit for Android and iOS applications, that serves as a runtime monitor for the use of private data. When data enters into Aegis, it is tagged with a privacy policy that controls access to the data. All computation is done inside the Aegis trusted execution environment and is checked in real-time for policy compliance. Applications never have direct access to the data unless the policy explicitly authorizes data to be released to the application. Policies are regular expressions that determine which operations are compliant. As operations are performed, both the data and policy react and transform accordingly. We have demonstrated the feasibility of Aegis by developing a location fetching application which anonymizes a location to a given radius bound.

Cognitive Science

Margaret Bigler ’22, Polyphony Bruna ’22, and Professor Janet Andrews (Cognitive Science)

Symposium Poster and Presentation Audio

Project Abstract

Learned categorical perception (CP) is a phenomenon where learning to place objects in categories influences how similar they appear, with objects in different categories becoming easier to tell apart and/or objects in the same category becoming harder to tell apart. Despite these effects being widely demonstrated, past studies exhibit low statistical power and the literature lacks a unifying theoretical framework. We seek to rectify these issues by conducting a systematic methodological investigation of learned CP, starting with replicating the effect under the conditions with which it has traditionally been reported, then exploring how successive methodological changes impact the presence of the effect. Our replication failed to show a pattern indicative of learned CP from comparing discrimination performance between a group that had learned a category distinction and a control group that had not. Through exploring our data to scrutinize possible key differences between our study and previous demonstrations of learned CP, we hypothesized that a combination of our stimuli being too easy to discriminate and the memorization of individual stimuli along each dimension obstructed the influence of category membership on discrimination behavior and was responsible for the absence of CP effects. We addressed this issue by lowering the discriminability of stimulus pairs and by increasing the number of stimuli in each category. Preliminary results suggest a possible learned CP effect and we plan to collect additional data to clarify the nature of the pattern.

Spencer Lee ’22, Ling Qi ’22 and Professor Josh de Leeuw (Cognitive Science)

Symposium Poster and Presentation Audio

Project Abstract

This project aims to build a suite of online games that replicate psychological experiments. In contrast with a traditional laboratory approach, conducting experiments via online games unlocks access to a larger and more diverse sample of participants. With a large sample, we are able to employ “radical randomization” as a strategy to assess the generalizability of findings. Radical randomization involves not only randomizing the independent variables key to the experimental question, which usually correspond to variations in the game’s level of difficulty, but also randomizing the seemingly less significant but potentially impactful variables, such as the color of a stimulus. This strategy allows us to explore more of the experimental design landscape and avoid basing findings on coincidental “sweet spots” of experimental features.

This summer, we developed two more games, assessing working memory and motor sequence learning. With the increase in size of our game suite, we developed a “party mode” in which friends can join an online game room, compete against each other in a series of games, and receive rankings among themselves. Apart from engaging more participants, this multiplayer setting also provides more flexibility for radical randomization. In traditional game design, games get more difficult as a player acquires more experience. In party mode, players are competing against each other, freeing us to select random levels every time a group plays.

Ivy Chen ’21, Eden Forbes ’21, Abigail Jenkins ’22, Hero Liu ’22, and Professor Josh de Leeuw (Cognitive Science), Professor Hadley Bergstrom (Psychological Science), Professor Bojana Zupan (Psychological Science), Professor Lori Newman (Psychological Science)

Symposium Poster and Presentation Audio

Project Abstract

In trying to identify and understand what the neural underpinnings are for an observed behavior, one of the critical challenges is recording neural activity in real time and space. Here at Vassar, we have recently developed a miniaturized microscope (i.e. Miniscope) that allows for imaging of the brain in a free-moving animal at single cell resolution. However, this process produces video data that is challenging to analyze due to both the noisiness of the data and the massive amount of data produced. We constructed a computational architecture to address these challenges and yield useful data from the Miniscope in a comprehensive and efficient manner. Furthermore, we implemented other analysis tools to analyze behavioral video data and in turn allow for the synthesis of a recorded behavior and its corresponding neural activity. Our architecture is derived from a series of open-source packages in Python, most importantly the Minian package from the Denise Cai lab at Mount Sinai and DeepLabCut. Additionally, the pipeline is configured to run on Vassar’s computer cluster, Hopper, to allow for efficient processing available to the whole college. In sum, the set of analytic tools amalgamated here allows us to derive neural signal and behavior from high-dimensional video data and opens the door for the addition of further tools to better decode the relationships between brain activity and behavior.

Ivy Chen ’21, Eden Forbes ’21, Abigail Jenkins ’22, Hero Liu ’22, and Professor Josh de Leeuw (Cognitive Science), Professor Hadley Bergstrom (Psychological Science), Professor Bojana Zupan (Psychological Science), Professor Lori Newman (Psychological Science)

Symposium Poster and Presentation Audio

Project Abstract

Functionality of the infralimbic (IL) subregion of the medial prefrontal cortex is required for aversive memory extinction. However, how groups of neurons (ensembles) in the IL respond over the course of aversive learning and extinction is unknown. This question is important because it is thought that coordinated ensemble activity underlies learning and memory. To address this question, a fluorescent miniaturized microscope (i.e., “the miniscope”) combined with a genetically encoded calcium sensor (GCaMP) transfected in the IL was used to record changes in calcium signaling dynamics at cellular resolution during aversive learning and extinction. With the acquired data, we are currently in the process of examining coordinated patterns of calcium signaling as well as freezing behavior using the open-source data analysis packages Minian and DeepLabCut, respectively. Frames gathered from miniscope imaging during the experiment will be used to track patterns of calcium signals from individual cells over the course of an entire behavioral session and across days using the Minian pipeline. From the behavioral videos, eight points (snout, left ear, right ear, middle of spine, base of tail, two points on tail, and tip of tail) on the mouse were labeled using DeepLabCut to create a skeleton which follows the animal’s movements based on a grid system. This information will be used to classify types of behavior (specifically freezing patterns) based on the body orientation and the velocity of movement. We are tracking changes in this behavior along with the neural activity over the course of fear conditioning and three days of conditioned stimulus extinction. These data will enable the analysis of ensemble activity in the IL while the animal learns and extinguishes an associative aversive stimulus.

Michael Jaklitsch ’21, Lingxiu Zhang ’21, Jason Han ’23, and Professor Kenneth Livingston (Cognitive Science)

Symposium Poster and Presentation Audio

Project Abstract

We use simulated robots to test hypotheses about basic principles governing the evolution of intelligent agents. Specifically, we hypothesize that mechanisms of gene duplication and subsequent differentiation increase the modularity of a robotic agent’s sensor/neural network/motor system over generational time, resulting in populations that are more evolvable, that is, better able to survive and even prosper when fitness landscapes shift. This summer we began the process of building a simulation environment, realistic robotic models, and the complex network of software necessary to study the evolution of populations of these robots across hundreds of generations. This summer’s work focused on the three core components of the software system, which will be optimized to run on Vassar’s multicore supercluster, Hopper. First, we designed and coded a genotype to phenotype (G-to-P) mapping algorithm that uses an unusual process-oriented (rather than part descriptive) mechanism to turn any randomly generated robot genotype into its corresponding phenotypic embodiment of sensors, motors, and neural network. This G-to-P mapping scheme is designed to allow simulation of gene duplication events as well as the mutation events typical in evolutionary robotics simulations. Next, we used the Gazebo simulation environment, which implements the Bullet physics engine, to construct a representation of the robot specified by the genome as well as the environment in which it must operate. Finally, we developed a fitness function that requires the solution to a behavior problem logically equivalent to the exclusive OR, a problem known to require neural networks of more than two layers to solve. Next steps will include the construction of the software infrastructure necessary to create a population of genotypes that will then be evaluated in the simulated environment and allowed to reproduce based on their success in that environment. A comparison of the evolvability of two populations, one evolved with and the other without the gene duplication mechanism turned on, will allow a clear test of our hypothesis.

Mathematics and Statistics

Mia DeStefano ’22, Wyatt Milgrim ’23, Ceci Villaseñor ’22, Emily Wadholm ’23, and Professor Adam Lowrance (Mathematics and Statistics)

Symposium Poster and Presentation Audio

Project Abstract

Imagine you tie a string into a knot and glue the ends together: this is a mathematical knot in 3-space. When the 3-dimensional knot is projected into the plane, we obtain a picture of the knot called a knot diagram. We can look at each crossing of the knot and determine which strand goes over and which strand goes under in the diagram. Start at an arbitrary place in the knot and follow the string around. If it alternates between going over and under at each crossing, then we say the knot is alternating. Our work is on a specific class of knots, called almost alternating knots. These are knots that have a diagram where one can switch a single crossing from over to under to create an alternating diagram. There are certain properties of knots called invariants which help to differentiate between distinct knots. Our main result uses an algebraic knot invariant, called the Khovanov homology of the knot, to test whether the knot is almost alternating. We also examine the 4-genus of a knot—the fewest number of holes in any surface in four-space that is bounded by the knot. We prove that there are upper and lower bounds on the 4-genus of many almost alternating knots; these upper and lower bounds differ by one.

Physics and Astronomy

Symposium Poster and Presentation Audio

Project Abstract

We used Matlab to simulate experiments in which a thin metal film of known material properties is grown on top of a thin sample layer and an ultrafast laser pump-probe method is used to heat and measure the thermal conductivity of the sample layer. We became familiar with a Matlab program published by researchers at The University of Illinois that simulated the experiment and used this simulation to determine the ideal experimental parameters for measuring thermal conductivity of MoSe2 and other novel materials that are of interest for nanoscale electronic and optoelectronic devices. A subset of the thermal conductivity simulation allows for testing the sensitivity of material parameters of the metal film, sample layer, and the substrate. We tested the sensitivity of the parameters at different laser modulation frequencies and pump and probe laser spot sizes. The sensitivity simulations allowed us to design the ideal experimental set up in order to optimize the experiment for measuring the thermal conductivity of the sample layer. For MoSe2, we have found that the ideal experimental set up is 1 m MoSe2 with 80 nm Al film, pump and probe spot size radius of 5 um, and 2.0 MHz modulation frequency. We have also designed other experimental set ups for samples <1 m of MoSe2.

Symposium Poster and Presentation Audio

Project Abstract

We investigated using an ultrafast pump-probe experiment to image samples by monitoring their optical properties. This process uses a pump laser pulse to excite a rapid thermal change in the sample, creating acoustic pulses in the substrate below. After a certain time, a probe laser pulse measures the change in optical properties at the sample’s surface. Thus far, we have used Aluminum as a transducer to create an acoustic pulse in the sample and to detect the pulse. One of the disadvantages of using Al as a transducer is that it creates low resolution images due to its low sensitivity to strain. Therefore, we conducted literature research to explore the possibility of using a Transition Metal Dichalcogenide (TMD) as the detecting layer in our pump-probe experiment. We were interested in which optical properties (reflectivity or transmissivity) to measure, how many layers of TMD to use, and which TMD (WSe2 or MoS2) would be most effective. We also investigated which phonon modes would get the strongest signal and whether surface acoustic waves could be used to improve image resolution. We found that TMDs would likely act as effective strain sensors due to their high sensitivity to strain. However, further experimental research is needed to make a conclusive statement.

Symposium Poster and Presentation Audio

Project Abstract

We have run computer simulations of an ultrafast pump-probe laser experiment that generates and detects surface acoustic waves (SAW) in periodic structures at the nanoscale. In these experiments, a ‘pump’ beam of light causes strain and generates surface waves in a structure by heating the surface and causing thermal expansion, and a ‘probe’ beam of light is used to measure the change in reflectivity of the structure, which can be used to identify surface waves with frequencies up to 50 GHz (the highest SAW frequencies that can be detected). Our simulations consisted of two parts: a mechanical simulation that uses the elasticities and densities of the structure to calculate and show strains in the structure, and an electromagnetic simulation that uses the strains to simulate changes in the optical reflectivity of the structure. We focused on how different polarizations of the probe beam affect the changes in reflectivity at both 400 nm and 800 nm wavelengths of light. Our simulations predicted significant differences in reflectivity between polarizations for some structures, and no significant differences in other structures. We compared our simulations to published experiments and found surface waves at similar frequencies as were found in experimental data, but polarization and wavelength dependence do not consistently agree with experimental data.

Project Abstract

Plasmonic nanoparticles are small metal particles, usually with diameters from tens to hundreds of nanometers. They have unique properties having to do with their absorption of light in the visible or UV range which has applications in medicine and in light-speed information processing with metatronics. As light interacts with the nanoparticles, the distribution of charge varies, and the response of the particle can be modelled as a lumped-element circuit. This model is well understood for single nanoparticles larger than 20nm, and for sets of nanoparticles if the gaps between them are larger than 3nm, but as they get small or close together, quantum effects are no longer negligible, and the circuit model fails. In this project, we sought to adapt the circuit model (a classical model) to predict these quantum effects. We focused on small nanoparticles, rather than coupled nanoparticles, and have defined a new circuit to model the quantum effects. This new circuit fits experimental data to the extent of having the proper resonance frequencies and Q-factors. However, more research is needed to determine if the parameters in these circuits have appropriate values for the physical situation.

Adam Moses ’21, Hannah Stickler, Wellesley ’22, and Professor Colette Salyk (Physics)

Symposium Poster and Presentation Audio

Project Abstract

Protoplanetary disks are the dusty feeding grounds around young stars out of which planets coalesce and grow. However, their small angular sizes and dustiness present challenges for detecting and studying planets embedded within them. One means of probing the hidden properties of these disks is via spectroscopy of rovibrational emission lines of carbon monoxide (CO), whose characteristics can be an indication of an embedded planet (e.g., Brittain et al. 2014; Regály et al. 2010, 2014). We present data and analysis of a number of 12C16O v=1-0 and v=2-1 emission lines originating from the protoplanetary disks around the young stars AS 205 N and AB Aurigae with the goal of identifying variability, possibly due to planets in the disks. The data from AS 205 N clearly show emission that decreases in intensity with time. The decrease occurs primarily for molecules with lower speeds, which correspond to greater orbital distances. This phenomenon could be due either to a decrease in emitting surface area (found to be about 15%), a decrease in temperature, or some combination of the two. The emission lines from AB Aurigae showed no significant variability; and had relatively low flux values.

Symposium Poster and Presentation Audio

Project Abstract

Due to the recent crisis of COVID-19, we decided to create a substitute or complement for introductory physics laboratory classes (College or HS). We have created a series of video-based graphical analysis questions that focus on concepts taught in mechanics with clear common misconceptions addressed. The videos we used are of real world motion such as someone shooting a basketball, billiard ball collisions, uniform and non uniform circular motion of an exercise bike wheel. These activities address concepts such as conservation of energy, simple harmonic motion, angular momentum, friction and forces to name a few. Videos were analyzed previously using Logger Pro software and each activity has the video shot, tracked motion images, graphs of relevant quantities and physics concept questions that rely primarily on getting information from the graphs. The questions in this compilation focus on developing student skills such as estimation, analyzing real world data and graphical interpretation. Each question has specific feedback for common wrong answers and has the correct answer with a supplemental explanation available. The video series will be implemented in the Expert TA system this fall.

Psychological Science

Symposium Poster and Presentation Audio

Project Abstract

Astrocytes have an important role in cognition as they possess a unique function to recycle glutamate (an excitatory neurotransmitter) from synapses using the enzyme, glutamine synthetase (GS). Dysfunctional astrocytes found in neurodegenerative disorders, like Alzheimer’s disease, have decreased levels of GS correlating with cognitive impairment. We seek to understand how GS levels relate to sustained attention within male and female Long Evans rats. Using the sustained attention task (SAT) created by McGaughy & Sarter (1994), the rats are tasked with discriminating between a signal trial (500 ms, 100 ms, 25 ms) and a non-signal trial. To examine the effects of time on task, we divide each session into three blocks of trials. Based on their performance on the task we divided our samples into groups of attending (higher performance) and non-attending animals. Our SAT analysis showed a significant difference between signal lengths, with 500 ms signal trials having the fastest reaction times and the most accurate results, and 25 ms signal trials having the slowest reaction times and the least accurate results. Between the signal and the nonsignal trial, the animals demonstrated a significantly faster reaction to signal trials compared to the nonsignal trial. Our SAT analysis also showed no significant sex difference between females and males, but showed that attending animals respond significantly faster than non-attending animals. There was a significant effect of task progression on attention, with accuracy of correct hits decreasing as the blocks increased. This demonstrates that the animals experienced a vigilance decrement as time progressed. Our data highlights the relationship between the reaction times of the SAT and attention. Additionally using immunohistochemistry for GS, we observed significantly higher levels of GS present in the prefrontal cortex and hippocampus of attending animals compared to non-attending animals. Thus we can further connect our understanding of attention to GS recycling. By understanding this connection between astrocytes and cognitive function, we can design therapeutic targets for patients suffering from disorders.

Annie Xu ’22, Mohtad F. Allawala ’23 and Professors Sue Trumbetta (Psychological Science), Adam Brown, and Maria Höhn (History)

Symposium Poster and Presentation Audio

Project Abstract

The concept of self-efficacy (SE) refers to an individual’s belief in their ability to perform the necessary actions to manage particular situations and reflects confidence in the ability to exert control over one’s own motivation, behavior, and social environment (Bandura, 1977). Self-efficacy has become an increasingly important construct in light of the COVID-19 pandemic. While the virus has a direct impact on the body, its repercussions, including the extended period of self-isolation that has ensued, can negatively affect an individual's psychological wellbeing and everyday functioning. Perceived SE regulates stress and anxiety arousal both physiologically and cognitively. People with a higher level of SE take bolder actions dealing with problematic situations while maintaining a lower level of physiological activation (Bandura, Blanchard, & Ritter, 1969). Furthermore, Brown and colleagues (2012) discovered the benefits of SE interventions in reducing intrusive recollections of aversive events, which in turn assist post-traumatic recovery. The present research focused on introducing smartphone technology as an immediate site for scalable, low-cost psychological support. The SeApp induced participants’ SE by prompting recollection of mastery experience, which is one of the four tenets of SE according to Bandura’s theory. Mastery experiences refer to times when an individual takes on a challenge and succeeds. The study aims to test the app’s feasibility and to investigate how daily SE and motivation can be enhanced, and perceived stress can be reduced, through daily memory-based training via a one-week intervention through the smartphone app. We are also interested in exploring the impacts of this SE intervention in a multicultural context. By recruiting both domestic and international students, we expect to have a better understanding of how populations from different cultural backgrounds benefit from recalling memories of personal success.

Symposium Poster and Presentation Audio

Project Abstract

One of the most consistently observed effects of chronic stress on the hippocampus, mediated through high levels of systemic glucocorticoids, is structural alterations of the branching of pyramidal cell dendrites. Data suggests that exercise can normalize many effects of stress, including dendritic restructuring (Trinchero et al., 2019). Interestingly, we have previously found that postnatal maternal exercise can also ameliorate many behavioral effects of stress in her adult offspring. In order to assess whether increased resiliency to chronic stress in runner-dam offspring is associated with structural changes in the hippocampus, male mice born from dams that either had post-parturition access to a running wheel or a standard housing cage underwent a chronic unpredictable stress (CUS) paradigm for 21 days. Following CUS, mice were perfused, and brains were Golgi stained to study dendritic arborization in the CA3 region of the dorsal and ventral hippocampus.

For this project, Golgi-stained neurons were visualized under a brightfield light microscope, and 3D reconstructions of neurons of interest were created using Neurolucida. The process included delineating and reconstructing the axon, dendrites, and soma of a neuron, thereby creating a digital, geometric model of the cell. The 3D reconstructions were used to visualize and morphometrically analyze several intricate neuronal structures. A currently ongoing Sholl analysis will reveal the dendritic length and the number of dendritic intersections that occur at fixed distances from the soma in concentric spheres, which will allow us to determine whether postnatal maternal exercise can rescue dendritic arborization from the effects of chronic stress. If structural differences are found between our groups, we will assess gene and protein expression patterns in these brain regions to probe for associated molecular changes which may drive the observed dendritic restructuring.