2022 Projects


New York City’s (NYC) water is considered the “champagne of tap water” due to its natural taste and clarity without the need for filtration. However, NYC’s search for water has left a legacy of displacement and destruction in its wake. Nineteen artificial reservoirs and controlled lakes, located up to 125 miles from Manhattan’s center, feed the system. The costs of creating and maintaining these reservoirs includes dozens of demolished communities. The goal of this research was to map NYC's impacts to reservoir communities through documenting their demolished buildings. The total number of demolished buildings, never before calculated, appears to be 1,127 submerged by reservoir waters and 793 on nearby lands. Using ArcGIS Pro and historic maps, the previously existing buildings were mapped based on location in the reservoir water line or in NYC’s Department of Environmental protection (DEP) property lines. With these draft maps, fieldwork was used to pinpoint building foundations which can increase map accuracy and building counts. Witnessing the total number of buildings demolished decenters NYC within the history of their water supply and contradicts the idea that NYC water comes from rural landscapes that are not otherwise used.


Translation is an essential biological process in which the sequence of a messenger RNA molecule is transformed into a functional protein. Initiation is the first and most highly regulated stage of translation, and involves one of the most important readings of the genetic code. Translation initiation sets the reading frame of mRNA translation and ultimately controls protein synthesis. eIF3 is a multi-subunit eukaryotic initiation factor, which is a protein complex that assists in the recruitment of ribosomes to mRNA during translation initiation. eIF3 participates in each phase of the translation pathway, promoting mRNA attachment and start codon recognition. Altered subunit expression of eIF3 has been linked to the development of a variety of human cancers. We are investigating the mechanistic function of the eIF3 complex in translation initiation through analysis of translational defects caused by deletion of multiple eIF3 subunits. By providing insight into the function of eIF3 subunits, as well as the role that individual subunits may have in translational impairment, we hope to gain a more complete understanding of eIF3 complex functionality, as well as to provide insight into the ways that eIF3 disruption can contribute to disease development. Our analysis has focused on three eIF3 subunit mutations: eIF3a/b degron, eIF3i DDKK, and Moe1. The degron mutation results in the deletion of the entire eIF3 subunit, and the DDKK mutation results in the loss of the eIF3i and eIF3g subunits. The Moe1 mutation causes deletion of the eIF3d subunit in fission yeast. We have accomplished translational analysis of all wild type and mutant strains using ribosome profiling, which provides insight into the global translational efficiency of the complete set of mRNAs across an organism’s transcriptome. We are also utilizing RT-qPCR analysis in order to analyze the translational capabilities of the degron and DDKK mutations in comparison to wild type strains, which will provide insight into the specific means of translational impairment caused by disruption of eIF3.

Translation, the second core event of gene expression, during which RNA is converted into protein, is a highly regulated and fundamental biological process that creates the essential molecules needed to perform molecular functions in the cell. Translation initiation is the most regulated and the rate-limiting step of translation. Dysregulation of translation initiation can cause cells to become cancerous and is observed in neurodegenerative disorders and infection. Eukaryotic translation begins with the formation of a pre-initiation complex (PIC), where eukaryotic initiation factors (eIFs) and the initiator tRNA assemble on the small ribosomal 40S subunit and scan the mRNA to identify the AUG start codon. eIF3, the largest and most complex initiation factor, wraps around the PIC and has roles in PIC formation, mRNA recruitment, and start-codon recognition. Previous ribosome profiling work in our lab has identified mRNAs whose translation is sensitive to mutations in the mRNA entry channel (mEnC) arm of eIF3 or degradation of the entire eIF3 complex. Using an in vitro reconstituted system with eIFs and ribosomal subunits purified from Saccharomyces cerevisiae, we are investigating the kinetics and efficiency of recruitment of these mRNAs with both wild-type and mutant eIF3. By identifying characteristics in the 5’ untranslated regions of these mRNAs that make them sensitive to mutations in eIF3, we will illuminate the molecular mechanisms by which eIF3 and its component subunits assist in mRNA recruitment. Additionally, we are also using firefly luciferase reporter constructs to monitor the expression of these mRNAs in extracts prepared from yeast strains carrying these eIF3 mutations. These experiments will shed light on the mechanisms of translation initiation and mRNA recruitment, allowing us to understand how defects in eIF3 contribute to disease progression.

Translation of the genetic code into protein is among the most essential cellular processes, without which life would not exist. Translation, and translation initiation, the rate limiting step of the whole process, are therefore highly regulated, and translational misregulation is linked to the proliferation of a number of diseases, including cancers, viral infections, and neurodegenerative diseases. Essential to initiation is eukaryotic translation initiation factor 3 (eIF3), which is primarily responsible for the formation of the pre-initiation complex, an assembly of eIFs and tRNA on the 40S ribosomal subunit, which binds mRNA and initiates protein expression. Understanding how eIF3 functions therefore allows us to study how its dysfunction is linked to aberrant translation. eIF3 subunits can not be purified separately from S. cerevisiae, our model organism, as they are all essential. To avoid this issue, we plan to reconstitute in vitro eIF3 with subunits recombinantly expressed in E. coli and perform gel shift assays and ribosome profiling to determine if initiation occurs normally under various conditions. This allows us to express mutated variants of eIF3 and track the effects these mutations have on translational efficiency, specifically how these mutations affect PIC formation, binding, and mRNA recruitment. We expressed and purified native yeast eIF2 and 40S ribosomal subunits for use in gel shift assays that track mRNA recruitment to the PIC and the effects of eIF3 on this process. We are now exploring interactions between eIF3 and a 40S subunit variant containing a mutated latch protein at the mRNA entrance channel to shed light on the effects of this mutation on binding and translation initiation. In future, we plan to test eIF3 complexes missing certain subunits, which allows us to investigate the role of specific subunits in the overall process of translation initiation and how defects in parts of the complex can lead to the development of disease.

Ribosomes are organelles responsible for binding to messenger RNA (mRNA) and translating genetic sequences into proteins. Translation consists of recognizing the start (AUG) codon and reading the gene-coding sequences that assemble the chain of amino acids, forming proteins. Translation initiation is a highly regulated pathway that sets the reading frame for protein synthesis. It starts with the formation of a ribosomal pre-initiation complex (PIC), which contains the small ribosomal (40S) subunit loaded with several eukaryotic initiation factors (eIFs) and the methionylated initiator tRNA (Met-tRNAi). The PIC then binds at the 5’ end of an mRNA and scans through the 5’-untranslated region (5’-UTR) until reaching the start codon. The 5 major subunits of eIF3 in yeast S. cerevisiae form a complex that contribute to each core event during translation initiation, including the recruitment of mRNA to the PIC. eIF3 has also recently been shown to participate in translational regulation in higher eukaryotes.

The PIC’s mRNA entry channel arm allows mRNA binding and scanning when open, then scanning arrest and start-codon recognition when closed. Our project explores variant strains of yeast– DDKK, which has a mutation that impairs the entry channel arm, and Degron, which disrupts the entire eIF3 complex. We focus on the wildtype (Y162) and DDKK mutant (Y170) yeasts. Using prior ribosome profiling, we identify mRNAs that are most and least sensitive to the mutations. We transform the yeasts using Luciferase reporters with the same 5’-UTR as those mRNAs to see if the levels of translation are attributed to the 5’UTR or other regions of the mRNA. By comparing levels of protein expression in the wildtype and DDKK transformants, we dissect the effect of the DDKK mutation on mRNA recruitment, scanning, and start-codon recognition. The underperformance of Y170 transformants suggest that alterations to the PIC’s entry channel arm have debilitating effects on translation initiation.

Ungulate herbivory is known to have a strong impact on forest composition, including the leaf litter invertebrate and soil microbial communities. Excessive white-tailed deer (Odocoileus virginianus) browsing has caused alterations in plant understory growth in northeastern forests—diminished sapling growth has created a large divide between overstory and understory saplings. The alteration in leaf litter quantity and quality from herbivory indirectly affects leaf litter invertebrates and soil microbes. In this study, we investigate the effects of deer herbivory on sapling diversity, leaf litter invertebrates, and soil microbial activity. We hypothesize that excessive deer herbivory will reduce sapling abundance and richness. We also hypothesize that deer browsing will reduce litter invertebrate abundance leading to reduced nutrient availability. We surveyed deer plots in Millbrook, NY at the Cary Institute and the Rockefeller University Field Research Center which were established in 1992. The two locations vary in their deer management practices; the Cary Institute has employed a controlled doe hunting strategy since mid 1970's, while there is no management at the Rockefeller location. We used 8 sites with paired fenced/unfenced plots across the two locations. We surveyed sapling community composition, collected leaf litter samples and extracted invertebrates with the Berlese funnel method, and analyzed soil cores for nitrification and mineralization rates with a potassium chloride (KCl) extraction method. We found that sapling diversity was negatively affected by deer herbivory, while invertebrate diversity was unaffected for the one-time measurement. Nutrient cycling may be affected as well.

Deer population density has increased across the Northeastern United States over the past 30 years. Deer herbivory has different primary and secondary effects on forest communities, including shifts in decomposition rates through changes in plant litter composition or invertebrate community assemblage. This study explores the relationship between different levels of deer herbivory and leaf litter decomposition in temperate forests in the Hudson Valley. Sugar Maple and Red Oak senescent leaves were separately placed in litter bags of either small or large mesh sizes to allow for differing levels of invertebrate interaction. Small mesh excludes invertebrates to test their effect on decomposition. Bags were placed at three different sites in the Hudson Valley with varying degrees of herbivory; Teatown Preserve, Vassar Ecological Preserve, and Huyck Preserve. Each site has two sets of paired fenced and unfenced plots to test for deer impacts and all sites are dominated by a maple and oak forest canopy.

Litter bags deployed in November 2021 were collected after 6 months of decomposition and analyzed for mass loss, total carbon, and nitrogen. Soil cores were collected in June 2022 and studied for extractable nutrient ions, total carbon and nitrogen, pH, and water content. Litter invertebrates were collected at each plot to quantify abundance and richness. Other abiotic measures included total light flux at the forest floor and soil moisture levels. Our initial data show correlations between soil pH in mineral and organic soil horizons with invertebrate abundance. We found no significant relationship between mesh size and mass loss across sites. This study is ongoing and currently reflects the first 6 months of decomposition. Mass loss rates and soil abiotic and biotic variables will be re-assessed over the next year, providing further insight into deer ecological impacts over extended time periods.

The Hubbard Brook Experimental Forest spans 8700 acres within the White Mountains of New Hampshire where long-term ecological research has been undergone since 1954. To understand how wildlife is responding to changes in forest condition as a result of climate change, motion detecting cameras have been deployed across the valley in a roughly linear distribution. These cameras have run continuously since 2018 and the images they collect are analyzed to monitor a variety of species, most notably moose, a species which is highly cold adapted and may see its habitat range shift as a result of climate change. Through camera occupancy modeling, estimating moose presence within the forest allows us to determine if the species is reacting to the increasingly volatile climate. This data can be further used to determine how moose populations are responding not just in central New Hampshire, but may serve to inform the distribution of the animal within the greater Northeast.

While the camera data is vital in allowing us to observe patterns of distribution of the species, it cannot be used in isolation to inform which kinds of environmental variables most contribute to moose occupancy. By importing camera data along with geographic layers into ArcGIS Pro, we established spatial relationships between known moose presence and several environmental variables using the MaxEnt (Maximum Entropy) model. Through this analysis, we found that within Hubbard Brook there were several explanatory variables which appear to be correlated to moose occupancy: proximity to wetlands or beaver ponds, deer presence in a given area, and average number of unsuitably warm winter days. These correlations indicate several interesting patterns which implicate the future of the Hubbard Brook ecosystem, especially in regards to the lack of deer and moose cohabitation. Deer are carriers for parasites which are fatal for moose, and as the central New Hampshire climate continues to become more suitable for deer, this interspecies dynamic could become unsustainable for a viable moose population.

Orb-weaving spiders catch almost every kind of flying insect, except one type: moths. But one taxon of spiders, the Cyrtarachninae, has evolved to capture moths, and they appear to do so, in part, by making a glue that is very different from that of their orb-weaving ancestors. Given the importance of glue in capturing moths, we sought to understand the biochemical properties of moth-specialist aggregate glue droplets. We used Raman Spectroscopy to analyze the spatial distribution of the chemical composition of the glue droplets that have been spread out flat, as they would be during capture of moths. We analyzed moth-specialist species Cyrtarachne yunoharuensis, Cyrtarachne nagasakiensis, and Mastophora hutchinsoni and one generalist species, Leucauge venusta. For all four species, we analyzed three locations: capture thread, intermediate space, and droplet edge. For M. hutchinsoni, we analyzed an additional location of windlass specific to the glue droplet for this species. For C. yunoharuensis and C. nagasakiensis, we focused on two additional locations of tendrils and droplet core. For L. venusta, we also analyzed the droplet core. Samples were then reanalyzed after being washed in Nanopure water, meant to separate the water-soluble components, salts and small proteins, from the insoluble components, larger glycoproteins. We hypothesize that there is a distribution of glycoprotein in the core of the glue droplets and salts towards the edges for moth-specialist species, as opposed to the homogeneous distribution of components previously found in generalist glues. This distribution of components allows for a rapid drying of the glue as it spreads into moth wings, gluing down scales before they are able to be shed and the prey escapes. This research is supported by the National Science Foundation, project #2031962 to CD and JL.

The wings of a moth consist of an uneven follicle-covered surface lined with thousands of detachable scales. One of the many hypotheses in behavioral ecology is that these scales evolved as a defense mechanism against the glue droplets of orb-weaving spider webs. When a moth comes into contact with an orb-web, their scales easily flake off, effectively freeing them. As an exception to this norm, the subfamily Cyrtarachninae are a moth specialist species whose glue flows especially well on moth wings, penetrating the scaleline and gluing them to the prey’s base cuticle. We hope to understand the physics of this mechanism by modeling a moth’s wing; we are generating a series of comparative models with varying scale arrangements, surface topographies and pitting them against variable glue types for fluid analysis. In doing so, we hope to gain a comprehensive understanding of how spiders’ glue effectively spreads across moth surfaces, as well as understand the physical interactions between varying moth and spider species and their respective scales and glues. To create our models, we first use auto-fluorescence confocal microscopy to image Eusarca confusaria wing surfaces and individual scales. We then virtually reconstruct and analyze the acquired z-stacks using Fiji, before uploading them into Fusion 360 and combining our individual surface and scale components to create a fully formed 3D model. This research is supported by the National Science Foundation project, #2031962 to CD and JL.

Predation is one of the most common threats to survival. To mitigate predation risks, many species rely on anti-predator behaviors and social communication with surrounding con- and hetero-specific individuals about predator threats. Both natural and anthropogenic acoustic noise is known to disrupt these responses to predators, but the mechanism by which responses are depressed is not clear. Noise within the same sensory domain as the predation cue, such as acoustic anthropogenic noise, can overlap with the frequency range of an anti-predator vocalization and may reduce response via masking. However, noise across sensory domains may also contribute to an overall greater attentional load that results in distraction, such as acoustic noise distracting from the visual recognition of a predator. In our study, we sought to clarify the mechanism by which anthropogenic noise depresses anti-predator response. Using a repeated measures field experiment we recorded the vocalizations and approaches of black-capped chickadees (Poecile atricapillus), white-breasted nuthatches (Sitta carolinensis), and tufted titmice (Baeolophus bicolor) during exposure to both a visual predator model and acoustic predation stimuli with or without simultaneous exposure to anthropogenic noise. We then quantified the approaches and recorded vocalizations in the pre-, during-, and post-trial periods, and additionally quantified changes in call number, type, and note composition. We found that noise condition and trial period interacted to influence Black-Capped Chickadee alarm calling rate; chickadees decreased their calling in noise regardless of predator type, suggesting the underlying mechanism of distraction. However, White-Breasted Nuthatches produced more alarm calls when presented with an acoustic stimulus in quiet rather than in noise, suggesting the mechanism of masking.

Urbanization and habitat fragmentation has dramatically decreased global biodiversity of birds. With urbanization comes the construction of roadways, generating a high level of noise pollution. Habitats in proximity to these roadways have been found to have a reduced number of breeding birds and a change in species composition. The Vassar Preserve provides diverse habitats that support a range of different bird communities. On the Preserve I surveyed bird species and the number of individuals of each species at 12 wetland plots. Six of these were considered urban plots which were less than 100 meters from the south edge of the Preserve and had high levels of noise pollution. The six remaining rural plots were considered greater than 200 meters from the edge of the Preserve and had a lower level of noise pollution compared to urban plots. I recorded a total of 349 birds from 37 different species. Species were categorized as urban tolerant birds and rural preferring birds. The total number of individuals and species of each bird category were compared between urban and rural plots. There was a significant difference between the total number of rural individuals found at urban versus rural plots (t-test, p=0.027). This indicates that a subset of bird species on the Preserve are being significantly affected by noise pollution caused by adjacent roadways. To mitigate the impact of noise pollution, future tree planting effects could focus on replacing dead ash trees along the southern edge of the Preserve.

Open fields, also known as Successional Old Fields (SOF), are home to over 250 species of plants on the Preserve and provide important habitat for pollinators, birds, and other fauna. This habitat type is declining in New York due to an increased prevalence of woody plants (woody encroachment), invasive species, and other threats. The Preserve maintains about 40 acres of Successional Old Fields with an increased frequency of mowing (annually) since 2017. This summer we surveyed five old field plots to determine how species composition has changed since 2017. This was done by re-surveying 5 x 5 m plots in old fields, identifying all vegetation types inside the plots, and determining relative cover for the herbaceous, shrub, and vine strata. We determined that the abundance of woody species decreased significantly from 2017 to 2022, indicating a decline in woody encroachment. However, there was a significant increase in the proportion of invasive species, especially in the shrub stratum. This suggests that mowing, while an effective tool for reducing woody plants, has not been a successful management strategy for reducing the proportion of invasive species relative to natives. Overall, from 2017 to 2022 woody species have become less of a problem in old fields, but the proportion of invasive species has become higher, indicating that future monitoring will be important.

The Conservation Action Plan (CAP) is a tool used to inform the restoration and management of the Preserve. These efforts are crucial to our mission to protect the ecological diversity and educational value of the land. Furthermore, the CAP is a way to determine if our management efforts are working and if approaches need to be changed. The Preserve is categorized into conservation targets of different scales. The largest scale divides the Preserve into three priority management areas: Eastern Forested, Western Forested, and Central Open. These areas are then subdivided by hydrology (whether the land is wet or dry) and then by community type. To highlight how the CAP is implemented, I will use the Shallow Emergent Marsh (SEM) as an example community. SEMs face specific threats on the Preserve, including invasive species like Japanese stiltgrass (Microstegium vimineum). Some SEMs have an extremely high proportion (>50%) of this invasive species, so that they can no longer be considered an SEM. In order to mitigate these threats, the CAP outlines management strategies such as boot brushes, outreach events, and restoration of native species. To determine the effectiveness of these practices, we use recovery wheels. We collect data on flora, fauna, and other metrics that are then compiled into a recovery wheel and compared to a healthy reference ecosystem. From this data, I found that SEMs on the Preserve have less than half of the native species expected in a reference community. Since 2017, two characteristic plant species (red maple and cattail) have disappeared, but two new native plant species (marsh fern and canary grass) have appeared. In the future, Preserve management may need to be reassessed to ensure the persistence of Shallow Emergent Marshes.


Two-dimensional (2D) transition metal dichalcogenides (TMDs) are single-layered semiconductors composed of transitional metals (M) and chalcogen atoms (X) in the form of MX2. 2D TMDs have many optoelectronic and spintronic applications, including serving as the detecting layer to sense acoustic pulses in pump-probe experiments. The optoelectronic properties of 2D TMDs can be tuned for these applications by changing the nature and value of their band gaps via applying strain, alloying, and doping. The individual effects of alloying and applying strain on 2D TMDs have been well studied, but no research has investigated their combined effects. Focusing on group-VIB 2D TMDs, we aimed to determine how band gaps can be tuned through simultaneous applied strain and alloying. We also sought to find an alloy whose band gap is highly sensitive to strain, which is a characteristic that is optimal for pump-probe experiments. We examined the WS2xSe2(1-x) and MoxW(1-x)Se2 alloys, varying the S/Se and Mo/W concentrations and applying compressive and tensile strains. The alloys’ electronic structures were examined using the PBE and HSE exchange-correlation functionals. We found that applying strain decreased the band gap of the TMDs. Applying strain also changed the band gap’s nature from direct to indirect, although the crossing point for this transition depends on the alloy concentration. Pure WS2 was also identified as the material whose band gap was the most sensitive to strain.

Titanium dioxide (TiO2), a wide-gap semiconductor, exists in three polymorphs: rutile, anatase, and brookite. Rutile TiO2 is the most stable of the polymorphs and is of great interest due to its ability to convert solar energy to electric energy through the photovoltaic effect. TiO2 is commonly used as a semiconductor material in dye-sensitized solar cells (DSSCs), which are photovoltaic cells in which molecular dyes are paired with an inorganic semiconductor to convert solar energy into electrical power. Molecular absorbers are often bound to semiconducting nanoparticles using chemical substituents that serve as anchoring groups adsorbed on the surface of the metal oxide. Historically carboxylic acid has been frequently used as an anchor in DSSCs, but new anchoring groups are currently being explored, such as boric acid, which have exhibited improved power conversion efficiency. In this work we aim to assess the effectiveness of functionalized boronic acid, as compared to boric acid, as an anchor group between TiO2 and molecular dyes. Adsorption configurations, including both molecular and dissociative configurations, were constructed on two differently sized surface slabs. The strength of adsorption on the rutile TiO2 (110) surface was determined using DFT+U calculations with Grimme’s D3 dispersion corrections. Through our structural, electronic, and vibrational analysis of adsorbate configurations, we identified the most stable anchoring modes for each type of boronic acid and determined the effect of different functional groups on the adsorption strength.

Tau is one of the many microtubule-associated proteins that have a role in the stabilization and formation of the microtubules found in neuronal axons. However, tau can also self-associate into fibers and eventually neurofibrillary tangles which are positively correlated with neurodegenerative diseases such as Alzheimer’s. In order to better understand the mechanism of pathological tau aggregation, it is important to understand which variants of the protein aggregate most readily. It has been previously found that the C terminus of the tau protein inhibits polymerization. In this study, the tau protein with a C terminal deletion of the last 41 amino acids (Cdel) was purified and characterized. A fluorometric assay was used to measure the polymerization of Cdel when compared to wild type. Atomic force microscopy was used to interrogate the intermolecular interactions between tau proteins when bound to a mica surface as well as to analyze the growth and quantity of tau fibers over a six day period. It was found that Cdel adopts a less extended conformation and forms fibers that were not significantly different in length than wild type. However, contrary to the literature, this study found that Cdel polymerizes the same amount and at a similar rate as wild type. Future research will include repeat trials on all past experiments as well as with other Cdel variants in order to solidify these conclusions.

The microtubule-associated protein tau facilitates assembly, organization, and stability of microtubules found in neuronal axons. While phosphorylation of tau is a mechanism for microtubule regulation, hyperphosphorylation results in a loss of tau function from decreased microtubule binding affinity. Moreover, hyperphosphorylated tau is found in high proportions in paired-helical filaments, prominent in neurodegenerative disorders such as Alzheimer’s Disease. To investigate the effect of this phenomenon on the structural and behavioral properties of tau in vitro, the current study employed 7-Phos — a tau mutant with seven specific site pseudophosphorylation mutations — to mimic the characteristic changes found in hyperphosphorylation. In vivo, tau has six naturally occurring isoforms that vary in the length of the N-terminal projection domains and/or the number of repeating domains in the microtubule binding region. For this study, three isoforms of 7-phos, varying in projection domain length, were investigated. To confirm the purification of the isoforms, SDS-PAGE was used to identify an upward band shift, a feature of 7-Phos. A fluorescence assay was performed to compare the polymerization of the 7-Phos variants with their wild-type counterparts. Atomic force microscopy was used to measure the length of 2N4R 7-Phos tau polymers over several days and to analyze the intermolecular interactions between molecules of 2N4R 7-Phos. We determined that the two longer 7-Phos isoforms, 1N4R and 2N4R, polymerize as well as or less than their corresponding wild-type proteins while the shortest isoform, 0N4R, polymerizes more than its wild-type. The 2N4R 7-Phos variant also extends further into space and forms longer polymers than its wild-type. More research in the differences of structure and intermolecular interactions between 1N4R and 0N4R 7-Phos and their wild-type variants can further address the effects of hyperphosphorylation and its link to neurodegenerative disorders.

Tau is a microtubule associated protein found in neuronal axons that stabilizes and regulates microtubules. It has six naturally occurring isoforms, which differ in the number of repeating regions in its projection domain and binding domain. In particular, the projection domain is a natively unstructured region which controls spacing between microtubules and mediates polymerization of tau into aggregates. To further understand the structure and function of the projection domain, this research focuses on the projection domain mutant R5L, which has an arginine to leucine substitution at the fifth amino acid. R5L is of particular interest as it has been found in patients with frontotemporal dementia with Parkinsonism in chromosome 17 (FTDP-17) and progressive supranuclear palsy (PSP), diseases which are highly correlated with tau polymerization into neurofibrillary tangles. Interestingly, previous published works have shown varying levels of R5L polymerization as compared to wild type, so no conclusions have been made thus far. This project focused on 2N4R tau, the longest isoform, and compared R5L to wild type tau. Fluorimetry and AFM imaging were used to measure total polymerization and filament length. Atomic force microscopy (AFM) was also used to measure the length that the projection domain extends off the surface while bound to mica, a substrate that mimics microtubule binding due to its charge density. R5L was found to have less total polymerization, but a similar filament length and projection domain length as wild type. Further research will include replicating these results, as well as trials on shorter isoforms of tau, 1N4R R5L and 0N4R R5L, to compare to proteins with shorter projection domains.

Tau is a protein in the brain which regulates microtubule assembly in the axons of neurons. Tau aggregation is associated with many neurodegenerative disorders including Alzheimer's disease, but the details of the protein’s structure and aggregation process are not well understood. Tau is a natively unstructured protein that has eluded traditional protein structural characterization methods such as crystallography, but analysis with atomic force microscopy (AFM) offers a novel method to interrogate tau structure. When tau is adsorbed on the AFM cantilever tip and onto a flat mica surface, the N-terminal projection domains of the protein extend outwards to form polyelectrolyte brushes on both the tip and surface. The AFM measures the forces between the projection domains of the opposing layers of tau as the tip approaches the surface. This project tested two hypotheses for the variability in tau force curves taken during replicate measurements: differences in AFM tip radii and inconsistent layer structure of the tau after rinsing with high ionic strength buffer. Tip radii were measured by collecting images of a roughened titanium surface with high aspect ratio features and using the resultant image to construct a model of the tip. Based upon these measurements, differences in tip radii were not significant enough to explain the variability in our force curves. Rinsing tau layers with higher ionic strength buffers was investigated as a method to develop consistent layer structure and thus consistent force curves. Preliminary data shows that using a 500 mM salt wash as opposed to 100 mM may increase consistency. Going forward, an even higher concentration of salt will be tested to determine if that will achieve consistency in force curves. Achieving reproducibility in full length tau force curves is necessary in order to accurately compare full length tau force curves to force curves taken with other tau variants and mutants.

Tau is a microtubule-associated protein that regulates the formation of microtubules. In its filamentous form however, tau can form Neurofibrillary tangles (NFTs), which are characteristic of numerous neurodegenerative diseases such as Alzheimers disease. Studies have shown that NFTs are composed principally of Paired Helical Filaments (PHFs) along with other precursors of filamentous tau. To further characterize tau PHFs a procedure using Atomic Force Microscopy was developed through the use of existing literature. Tau polymerization was induced through the use of arachadonic acid and imaging was carried out using the AFM operating in intermittent contact mode. AFM analysis indicated an overall increasing average filament length as well as an increasing total filament length for up to 6 days of data collection. Statistical analysis was conducted on data that was gathered on both wild type tau and other tau variants. It is further implied by these results that certain morphological differences in the structure of tau play a role in the polymerization of the protein.

Polysaccharide degradation is an important part of digestion. Bacteria in the gut microbiome assist in digestion by breaking down large polysaccharides that are otherwise indigestible. Bacteroides thetaiotaomicron (B. theta) is an abundant bacteria in the microbiome and it utilizes the starch utilization system (Sus) to intake and degrade polysaccharides. Key components of the Sus system are the TonB-dependent transporter, TonB, ExbD, and ExbB proteins. ExbD assists in the movement of polysaccharides by using proton motive force to move TonB and the associated transporter by wrapping and pulling on TonB’s linker. In B. theta, many suspected ExbD homologs are present in the genome due to similarities between protein alignments of the regions of interest. In this research, the role of genes BT2052, 2053, 2666 and 2667 and whether they code for an ExbD protein are investigated by deleting these genes from B. theta’s genome utilizing allelic exchange methods and natural bacterial recombination. From a successful deletion of B. theta’s genes, growth patterns and differing morphology will be studied to determine the effect of the deletion and therefore the role of the encoded protein in starch utilization. Sequencing of plasmids designed to delete BT2666-2667 showed that the insert failed to stay intact during the deletion process, and efforts to increase effectiveness included designing new primers for the deletions and using different enzymes. With these alternative approaches, we have successfully generated plasmid inserts that will contribute to the assay of the proteins and their genes.

Bacterial microorganisms play an important role in regulating human health and digestion. The system by which these bacteria “eat” - or, take up complex sugars - has been extensively studied in Bacteroides, prevalent members of the gut microbial community. The gram-negative model organism Bacteroides thetaiotaomicron (B. theta) relies on specialized membrane proteins known as TonB Dependent Transporters (TBDTs) to move large sugars into the cell. As the name suggests, this transport is energized by TonB proteins. Previous research examined whether the TBDT specific to a starch uptake preferentially pairs with one or more of the 11 identified TonB genes in B. theta. Unexpectedly, deletion of BT3192 (TonB7) produced a lag in bacterial growth and gene complementation did not rescue the lag phenotype, suggesting that removing TonB7 from the genome disrupts some essential process(s) other than TonB interaction with the polysaccharide transporter. To achieve a ΔTonB7 strain without this phenotype, I redesigned the deletion scheme to instead target a highly conserved portion of the gene that is predicted to interact with the TBDT. Two such regions were identified—the entire C-terminal domain and a beta-sheet within that domain. I then attempted to construct a vector that will allow for deletion of each of these regions via allelic exchange in B. theta. When sequenced, these deletion plasmids did not contain the full DNA insert fragment, designed to facilitate recombination in B. theta. Troubleshooting the ligation of plasmid and insert had no effect on these vector constructs. Moving forward, I will use a different gene deletion protocol that has shown previous compatibility with the chosen plasmid. Once we can disrupt TonB7-TBDT interaction without the accompanying lag phenotype, we will better understand the impact of TonB7 on transporter activity.

Starch is digested in the human body by bacteria in the gut microbiome. Bacteroides thetaiotaomicron exists in the large intestine and helps the body break down starch via the starch utilization system (Sus). Within this outer membrane complex, SusC is designated as a TonB-dependent transporter which features a conserved motif known as the Ton Box. The starch molecule binds to SusC leading to a reaction between the Ton Box and the inner membrane protein TonB. This opens a chanel in SusC, allowing starch to come into the cell. There are 11 different TonBs that are a part of a larger intermembrane Ton complex. Our research is focused on the structure and function of TonB 2, 3, and 5 to better understand their role in starch transport. The goal is to express the TonB protein in Escherichia coli cells and then purify them using fast protein liquid chromatography (FPLC). We amplified an expression plasmid to create a linear pETite backbone and amplified the region encoding for the C-terminus of each TonB protein. We combined the two pieces of DNA and transformed them into E. coli. All three protein genes were successfully cloned. They were sent for sequencing to confirm the formation of the plasmid, and TonB 2 and 3 were successful. The plasmid was then successfully inserted into Rosetta pLysS cells. We used FPLC to attempt to purify the protein and found that TonB 2 and 3 were expressed solubly, but were not purified sufficiently. Future work includes using different protocols to obtain more purified protein which can be used to determine the protein structure with x-ray crystallography, and to understand the interaction between TonBs and SusC.

A synthesis of six prochiral imines substates and their asymmetric reduction by organometallic catalysis is presented. Substrates 1-3a-b were synthesized from the ketones 1-indanone, α-tetralone, and β-tetralone by heating them with benzylamine or p-tolylbenzylamine and a zinc catalyst in n-heptane to eliminate byproduct water. The use of n-heptane is an innovation which led to cleaner product and greater yield than using more toxic benzene or toluene; except for substrate 3a¬, they crystallized out of solution during concentration in vacuo with cooling. Iridium/H2 and hafnium or zirconium/BH4- catalysts were used to asymmetrically reduce the imines to amines with 85-99% conversion. The compounds were analyzed by X-ray diffraction, 1H NMR, 13C NMR, IR, GC/MS, and chiral GC with a Supelco β-DEX 325 column to investigate the enantiomeric excess. The crystal structures of substrates 1a (shown above), 1b, 2a, 2b and 3b were obtained via X-ray diffraction. Substrates 3a and 3b (right) obtained from β-tetralone, which has acidic protons, were obtained in the enamine form.

Cognitive Science

Learned Categorical Perception (LCP) is the phenomenon where learning to categorize a set of stimuli affects perceptual judgments of them. After learning, stimuli in different categories are easier to distinguish and/or stimuli in the same category are more difficult to distinguish. During last summer’s URSI, we used a set of 16 artificially generated faces as our stimuli to test for LCP effects. One study provided feedback after each trial in the discrimination task while another study did not; LCP effects were evident only in the condition without feedback. This summer we ran two more studies to further investigate this negative effect of feedback on LCP and to expand the stimulus set to make it more realistically variable. Random variation was added to the original set of 16 faces by “style-mixing” them with a randomly chosen set of 36 faces. Style-mixing adjusts small features of the face such as the skin texture, lighting, and hair color without impacting the structure of the face. This resulted in 36 sets of 16 faces for a total of 576 unique stimuli. We analyzed the data using sequential Bayes Factors with an informed prior based on our previous results and a justified preset stopping rule for ending data collection. Our results showed strong evidence of LCP for both the no-feedback and feedback versions, suggesting that LCP is robust to the presence of both feedback in the discrimination task and random variation in the stimuli. However, we have direct evidence that at least some other kinds of stimuli do not produce LCP and the strength of many published reports of LCP is questionable. We are therefore also preparing to conduct a meta-analysis on the entire body of relevant research to investigate the reliability of the LCP phenomenon.

Gaze prediction on a web browser is a powerful feature, with use cases stretching into easily distributable cognitive science experiments, accessibility features such as interacting with a browser, and many more. While a handful of software packages exist to solve this, they are not ideal for online experiments due to factors like a lack of spatial and temporal precision, along with the uniqueness of each user’s eyes and facial features. In order to alleviate these troubles, deep learning methods such as dimensionality reduction and neural networks are employed.

Firstly, we implement a variational autoencoder (VAE) model, which is used to create low dimensional representations of data, to encode facial features. This data, along with the relative location of the irises to the rest of the eyes, is then piped into another deep learning model that will learn to predict where a user is looking at the screen, regardless of head pose. To train and evaluate these models, data was collected from online participants that looked at 129 predetermined points on their screen. The data gathered from this experiment over time will not only help teach the deep learning model, but will also serve as a public use dataset that may also be used to train other models attempting to predict gaze. In addition, a developer-friendly free and open source JavaScript library was created in order to interface with the model’s predictions. The library stores and timestamps webcam video data through modern web browser functions that allow the developer to process through the model, along with eventually being able to support real-time and delayed gaze analysis. Finally, the library contains basic reusable components, expediting the process of camera access and calibration so that developers can focus on building the project that they need the gaze detection model for.

Creating a humanoid robot is a unique challenge in the field of robotics, even as it offers unique opportunities. Cognitive science theory includes acknowledgment of the role that the embodiment of an agent plays in determining the nature of its intelligence. Creating a humanoid robot means that one must take into account the constraints imposed by the way in which the human body appears and moves. This affects not only the hardware engineering of the robot but also creates many specific software control problems to be solved if the robot is to be a useful platform for the study of human perception-action systems. We adapted the open-source 3D body plan for the InMoov robot to construct a humanoid robot capable of many human-like actions, from simple behavioral responses to more complex action sequences and some low-level artificial intelligence including speech recognition and production and visually guided reaching and grasping. The platform will be continuously modified and updated as we learn more about the limitations and the potential of the prototype. The robot will serve as a demonstration platform in courses and as a research platform for the investigation of closed-loop perception-action systems and human-robot interaction patterns among other problems in cognitive science.

A Simple Temporal Network (STN) is a set of time points and constraints that can be used, for example, in automated scheduling. An STN can be represented as a weighted, directed graph, where the nodes represent time points and the edges represent constraints. An STN is consistent if it has a solution, which is a set of values for the time points that satisfy all of the constraints.

An STN can be represented by a matrix where each entry represents the weight of an edge. For space efficiency, we used the CSR (Compressed Sparse Row) Matrix representation.

We implemented a variety of algorithms for checking the consistency of an STN and generating solutions, including Floyd-Warshall (Floyd, 1962), Bellman-Ford (Bellman, 1958; Ford, 1956), Dijkstra (Dijkstra, 1956), and Yen (Yen, 1971). We used the Python/Cython programming framework which generates much faster code than ordinary Python. We applied the algorithms to both randomly generated STNs and STNs derived from a publicly available benchmark. On consistent STNs, both dense and sparse, the fast version of Bellman Ford (Bannister & Eppstein, 2012) was fastest. On inconsistent STNs, an alternative to Bellman Ford, suggested by Romeo Rizzi from the University of Verona and based on an algorithm by Ramalinam et al. (1995) was fastest.

We also implemented algorithms for converting STNs into dispatchable form, which enables efficient execution of STNs in real time. Our experiments confirmed that the FDM algorithm (Morris et al, 1998) is the fastest. Finally, we implemented and tested the BFCT algorithm (Cherkassky et al, 2009) which finds negative cycles in inconsistent STNs.

Overall, we have created a Python/Cython library of efficient algorithms for managing STNs that researchers can use, whether in the Temporal Reasoning Lab at Vassar or elsewhere.

Cloud computing provides an attractive model of remote computation, offloading the responsibility of managing hardware for an ever-growing base of clients. For efficiency and scalability, cloud providers use virtual machines (VMs) controlled by a Virtual Machine Monitor (VMM) to provide isolation and to allocate hardware resources among the client VMs. Unfortunately, under this model, clients need to trust that the VMM will not collect or leak any sensitive data. To combat this, computer manufacturers have released extensions to their processors to allow VMs to encrypt memory at the hardware level, protecting their data from the VMM and other VM clients. Our work looks at the Secure Encrypted Virtualization (SEV) processor extensions that AMD have introduced in their latest line of server hardware. Booting an encrypted client uses a specialized boot process that closely emulates the startup of a physical machine. We have characterized the performance of the SEV boot process, showing it is prohibitively slow for use cases like serverless computing, where minimally configured VMs (microVMs) are expected to boot in just tens of milliseconds. We introduce a new boot technique, called SEV Direct Boot, to start encrypted VMs faster, taking advantage of the fact that the system is already bootstrapped and removing unnecessary steps while maintaining the security guarantees provided by SEV. Our method allows encrypted VMs to boot quickly, closing the gap in startup times between encrypted and unencrypted VMs.

Computer Science

A Simple Temporal Network (STN) is a set of time points and constraints that can be used, for example, in automated scheduling. An STN can be represented as a weighted, directed graph, where the nodes represent time points and the edges represent constraints. An STN is consistent if it has a solution, which is a set of values for the time points that satisfy all of the constraints.

An STN can be represented by a matrix where each entry represents the weight of an edge. For space efficiency, we used the CSR (Compressed Sparse Row) Matrix representation.

We implemented a variety of algorithms for checking the consistency of an STN and generating solutions, including FloydWarshall (Floyd, 1962), Bellman-Ford (Bellman, 1958; Ford, 1956), Dijkstra (Dijkstra, 1956), and Yen (Yen, 1971). We used the Python/ Cython programming framework which generates much faster code than ordinary Python. We applied the algorithms to both randomly generated STNs and STNs derived from a publicly available benchmark. On consistent STNs, both dense and sparse, the fast version of Bellman Ford (Bannister & Eppstein, 2012) was fastest. On inconsistent STNs, an alternative to Bellman Ford, suggested by Romeo Rizzi from the University of Verona and based on an algorithm by Ramalinam et al. (1995) was fastest.

We also implemented algorithms for converting STNs into dispatchable form, which enables efficient execution of STNs in real time. Our experiments confirmed that the FDM algorithm (Morris et al, 1998) is the fastest. Finally, we implemented and tested the BFCT algorithm (Cherkassky et al, 2009) which finds negative cycles in inconsistent STNs.

Overall, we have created a Python/Cython library of efficient algorithms for managing STNs that researchers can use, whether in the Temporal Reasoning Lab at Vassar or elsewhere.

Cloud computing provides an attractive model of remote computation, offloading the responsibility of managing hardware for an ever-growing base of clients. For efficiency and scalability, cloud providers use virtual machines (VMs) controlled by a Virtual Machine Monitor (VMM) to provide isolation and to allocate hardware resources among the client VMs. Unfortunately, under this model, clients need to trust that the VMM will not collect or leak any sensitive data. To combat this, computer manufacturers have released extensions to their processors to allow VMs to encrypt memory at the hardware level, protecting their data from the VMM and other VM clients. Our work looks at the Secure Encrypted Virtualization (SEV) processor extensions that AMD have introduced in their latest line of server hardware. Booting an encrypted client uses a specialized boot process that closely emulates the startup of a physical machine. We have characterized the performance of the SEV boot process, showing it is prohibitively slow for use cases like serverless computing, where minimally configured VMs (microVMs) are expected to boot in just tens of milliseconds. We introduce a new boot technique, called SEV Direct Boot, to start encrypted VMs faster, taking advantage of the fact that the system is already bootstrapped and removing unnecessary steps while maintaining the security guarantees provided by SEV. Our method allows encrypted VMs to boot quickly, closing the gap in startup times between encrypted and unencrypted VMs.

Earth Science

Volcanism in NW Nicaragua is due to the subduction of the Cocos plate beneath the overriding Caribbean plate at rates of 70 – 90 mm/yr. The modern angle of subduction varies along the arc with the steepest angles beneath NW Nicaragua at 75°–80°. Faulting and structural evidence suggest that slab rollback has occurred, resulting in the westward migration of volcanic activity towards the subduction zone. Volcanism in Nicaragua is typified by basaltic-intermediate compositions from 100 Mya until ~ 7 Mya, when an apparent regional magmatic gap began. Typical arc activity resumed ~350 kya. Our study examined at least four distinct rhyolitic (70–75 wt% SiO2) eruptions that have been dated between 570 (±70) and 420 (±40) ka. Aside from a flare-up which ceased 7 Mya, rhyolites are not commonly erupted.

Thermodynamic calculations using the rhyolite MELTS v1.2 software package show that equilibrium crystallization conditions for these magmas are relatively cool (900–950 oC), hydrous, and occur at a depth of ~13–15 km. This finding is noteworthy since crustal thickness in the region is only 25–30 km and magma storage depths of the neighboring stratovolcanoes are typically < 10 km. At these depths water will be undersaturated, though our pumice clasts have a vesicularity of ~80 (±8) vol%. This suggests rapid syn-eruptive exsolution of water and other volatiles, which would drive dangerous, highly explosive eruptions. Pumice clasts contain various proportions of feldspar and pyroxene crystals with complex morphologies suggesting these magmas have experienced multiple, complicated differentiation paths.

Future studies will focus on plagioclase feldspar, a mineral ubiquitous in volcanic rocks which crystallizes in all compositions and develops textures reflective of formation conditions. We found crystal textures that indicated a variety of magma chamber cooling conditions and ascent patterns.


New York’s Climate Leadership and Community Protection Act (CLCPA, 2019) aims to cut statewide emissions to 85% below 1990-level GHG emissions by 2050. A key strategy to reach this goal is for each town, village, and county across the state to take an active role in reducing its emissions. A greenhouse gas (GHG) inventory is the first step in understanding a town’s emissions portfolio. A GHG inventory estimates emission from different sectors of the municipality—the town owned and operated sources—and the community at large—an estimate of emission totals from all residents within the community. We produced both a community and municipal GHG inventory for the Town of Kent, NY. Using guidance from NYSERDA along with ICLEI we calculated emission values for built environment (residential, commercial, industrial, process and product use), transportation, and waste. The municipal inventory was created using data collected from town offices and other local sources while the community inventory was based on consumption estimates. Municipal operation emissions constituted less than 1% of community emissions. Kent’s emissions were predominantly transportation (43%) and residential energy (42%), followed by waste (9%), process and product use (4.8%), and commercial energy (1.2%).

In terms of larger implications, we observed that it can be difficult for a municipality to undertake an inventory if they cannot afford outside consulting assistance. Our group began the process of developing tools which any concerned town member could utilize to conduct their own GHG inventory for their community without much prior knowledge or skills in GHG accounting.

Mathematics and Statistics

A knot is a simple closed curve in 3-space, and two knots are considered to be the same if one can be deformed into the other by stretching and bending it through 3-space. A knot is 2-bridge if the fewest number of maxima among any representative of the knot is two. The crossing number of a knot is the fewest number of transverse double points when any representative of the knot is projected to the plane. The genus of a surface is the number of holes in that surface. When we talk about the genus of a knot, we are talking about the fewest number of holes in any surface that the knot bounds. Our research focuses on the genus of 2-bridge knots with a fixed crossing number. Building off of the formula of the average genus of a knot that was found previously by Cohen and Lowrance, we compute statistical properties of the genus distribution of 2-bridge knots of a fixed crossing number. We do this by, first, finding a formula for the number of 2-bridge knots in a specific crossing number and genus, both recursively and explicitly. Then, using this formula, we can find other formulas for the median, mode, and variance of the genus distribution of 2-bridge knots of a fixed crossing number c. We also prove that the sequence of the number of 2-bridge knots of crossing number c and genus g is log-concave.


Bi2Te2Se and WSe2 are materials that can be grown so thin they can be called 2-D materials and they have applications in electronics and nanoscale devices. In particular, Bi2Te2Se is a relatively newly accessible material, therefore there is limited research available with no experimental results for the speed of sound or other elastic properties. In this project, mechanical exfoliation was used to deposit Bi2Te2Se onto silicon wafers that were pre-coated with varying layers of chromium and gold to improve bonding between the Bi2Te2Se and the substrate. An Atomic force microscope (AFM) was used to measure the thickness of the deposited flakes of Bi2Te2Se. Once the thickness was known, a pump-probe laser reflectivity experiment was used to find a speed of sound of 2060 ms-1 with a standard deviation of 36 ms-1. The pump beam of the laser induces strain waves in the crystal lattice causing changes in reflectivity of the crystal measured by the probe beam. The strain waves reflect off the sample’s surface, and return again so the signal produced by the strain waves can be used to deduce the speed of sound. The signal was reproduced through a MatLab simulation using the thickness of the flake determined by AFM to find the speed of sound. In a second experiment, the reflectivity setup was changed to measure fluctuations in transmission due to the strain waves on a transparent sample. The acoustic and electronic properties of WSe2 were studied at different laser wavelengths. Further research will look at Bi2Te2Se with the transmission set up and acquire more results for the speed of sound in Bi2Te2Se.

Rock samples collected in Dutchess County show evidence of metamorphism, transformation due to heat and pressure, as a result of the Taconic Orogeny, the mountain-building event that shaped the east coast of North America 400 million years ago. Erosion has since exposed rocks that were deep in the mountain range when it formed, and to understand the mountain-building process, we would like to know the conditions under which these rocks were created. Rocks consist of a set of minerals, which in turn consist of oxides whose composition can be analyzed. This set of minerals changes with pressure and temperature, so the minerals present in a given bulk composition constrain the conditions that the rock last experienced. For a fixed pressure, temperature, and bulk composition the observed set of minerals represents the minimum of the thermodynamic potential called Gibbs free energy. The software Perple_X uses databases of thermodynamic properties to calculate the set representing the minimum Gibbs free energy for a given bulk composition over a range of pressures and temperatures. The minerals in the calculated set and the minerals observed in the samples can be compared to provide the desired pressure and temperature constraints. To ensure the accuracy of the computational process, we first modeled samples analyzed by Whitney et. al. (1996), and compared our results to the reactions they observed. The models showed an unexpected excess of stable hematite (Fe2O3) which is dependent on the initial ratio of FeO to Fe2O3 in the bulk composition. A larger fraction of FeO in the initial composition yielded reactions closer to those expected, indicating more reducing conditions. Another complicating factor is that volatile oxides like H2O and CO2, the relative proportions of which can significantly change the computed set of minerals, cannot easily be measured. Further work is necessary to investigate the effect of different fluid ratios in these models.

We analyze the locomotion of the nematode C. elegans using optical diffraction. C. elegans is a microscopic worm of about 100 μm. These nematodes are commonly used as a simplified model for more complex organisms including humans. Due to their biological similarities, they are used to study bodily systems such as neuronal pathways. For locomotion analysis, the worm is placed in a quartz cuvette filled with distilled water, and allowed to thrash around. A laser is then directed at the live worm, producing a dynamic diffraction pattern. We constructed an optical setup that is used to refine the light beam to ensure symmetrical diffraction patterns. Temporal variations in light intensity at just one point in the diffraction pattern can reveal information about the entire worm’s locomotion. Once the dynamic diffraction pattern has been generated, it is captured using a CMOS camera, where a video of the diffraction pattern is recorded. We are then able to select specific pixels in the video, and generate a time series of the light intensity at a specific point as the worm thrashes. Once we obtain these time series, we can then perform computational analysis to determine if chaotic markers such as the Lyapunov Exponent are consistent at various points in the diffraction pattern. The presence of these chaotic markers, a positive Largest Lyapunov Exponent of 1.1406 1/s with an uncertainty of +/- 0.2 1/s indicates that the worm’s locomotion is likely to be chaotic and can lead to a greater understanding of neuronal dynamics.

Every day the scientific community gains a deeper understanding of how the human nervous system functions. The Vassar Applied Optics Laboratory is investigating the neuronal dynamics by analyzing the diffraction patterns of Caenorhabditis elegans in motion. With a completely mapped genome of only 302 neurons, C. elegans is a model organism for investigating the brain functions of more complex organisms. The biological insight into the nervous system dynamics of C. elegans is critical to inform the physical and computational components of this study by identifying motor neuron pathways and cataloging neuronal locomotory mutations. I was responsible for overseeing the cultivation and maintenance of the worm cultures and resolving any threats to the viability of our specimens, such as fungal infestations. The cultures were grown at 20 degrees Celsius in nematode growth medium agar petri plates, using C. elegans strain N2 and Escherichia coli strain OP50 as a food source. Maintenance included conducting a plate transfer every 4 to 5 days to allow consistent data collection with specimens in the same larval stage. The goal of this research team is to analyze, model, and predict C. elegans motor function as a means of understanding how these nervous system elements function together as a network. In the continuation of this research, this collected archive of information will allow us to explore the neuronal defects in mutants, and inform our researchers in an interdisciplinary understanding of our results.

We forecast the fluctuations in the dynamic diffraction pattern of C. elegans using artificial neural networks and reservoir computers. The time series data has a Lyapunov time of roughly 1 second. We aim to approach this prediction limit using machine learning. We train a multilayer perceptron network (MLP) with three hidden layers, each containing ten neurons, against the time series data from 2019. The MLP is not able to embed the complex behaviors of the 2019 data when making long-term forecasts. Recurrent networks are a far more attractive solution as we do not know how or when a feature in the time series will impact future behavior. An echo state network is able to embed more of the complexities of the data. The ESN still falls short when it comes to long term forecasts, however over time, single step predictions become exponentially less accurate at a rate that appears to be bounded by the previously established largest Lyapunov exponent observed for the locomotion of C. elegans. We plan to continue to tune our reservoir computer to establish confidence in the error trend being bounded by the LLE, as well as explore other means of forecasting the time series.

Many mathematical models have captured the action potential function of the central pattern generator that causes neural signal-passing. James Keener’s 1983 paper established the application of one of these models—the FitzHugh-Nagumo model—to an analog electrical circuit. Building upon Keener’s work, we sought to create electrical equivalent circuits for motor neurons of the nematode Caenorhabditis elegans, ultimately hoping to couple the circuits in order to pass signals controlling the locomotion of an electronic C. elegans. Our research this summer focused on creating singular circuits that replicated the Keener model. We began by using CircuitLab, an online circuit simulator. However, we found several limitations to the simulation, which encouraged us to move on to the physical circuit despite failing to obtain the exact same waveform as Keener. For the analog circuit, we created two versions, one based on Keener’s paper and the other based on previous work of one of our coauthors, Anshul Singhvi. From both, we obtained a waveform very similar to that of Keener, but not an exact replica. Adjusting the values of capacitors and resistors within the circuit also failed to produce the desired waveform. Further work should continue to examine the circuits and their relation to the biological neuron.

The microscopic nematode C. elegans has a simple nervous system that can help researchers better understand how neuronal dynamics function. We observe worm locomotion using an optical setup that generates a laser diffraction pattern. Optical diffraction is advantageous because it can resolve subtle changes in motion across multiple scales with high precision, unlike an optical microscope. While the optical setup illuminates the worm, its dynamic diffraction pattern is recorded using a CMOS camera. Time series are created from the recorded diffraction pattern by finding the intensity at a single point in each frame. These time series can then be analyzed for chaotic markers such as the Largest Lyapunov Exponent, determined through delay coordinate embedding and using Rosenstein’s algorithm. Delay coordinate embedding allows us to use the single variable data to recreate a representative version in the multivariable phase space of the system. We explored multiple parameters in the algorithm such as the lag and embedding dimension to determine which parameters unfold the locomotion in phase space. We found positive Largest Lyapunov Exponents of around 1.1406 1/s (with uncertainty of ± 0.2 1/s) for these time series, showing that this dynamical system is likely to be chaotic. This analysis of experimental time series can be compared with established neuronal models to gauge the accuracy of our current understanding.

We present the implementation of a near-field scanning optical microscope (NSOM) to observe the magnetic field generated by a set of coils. While previous instruments have been developed that can measure these fields at the nanoscale, they are capable of differentiating only one field component at a time. Our device, however, is capable of three-dimensional measurement simultaneously at the macro scale using orthogonally positioned magnetic probes based on induction coils. This novel approach enables the possibility of mapping electromagnetic fields by using Maxwell’s equations to derive the electric field associated with the measured magnetic field. Over the course of our study, we mapped both circular and polygonal loops of wire at 185 kHz, which matched the resonant frequency of our magnetic probes. The map generated from the circular loop closely followed the expected theoretical result, while the polygonal loops presented richer opportunities for analysis due to both their asymmetries and the presence of vertices. While the theoretical expectation would be a field of measurably differing magnitude at each vertex, we instead saw magnitudes more akin to those present along the length of each side. Further investigation is required to determine what factors could be contributing to this result, though we hypothesize a number of possible factors including but not limited to insufficient resolution, in terms of both sensitivity to the field and scan resolution, or imprecision in loop construction.

The relationship between the properties of protoplanetary disks, such as their chemical composition and substructures, and the exoplanets that are formed from them is not yet well understood. Previous work on this subject has been completed by the Disk Substructures at High Angular Resolution (DSHARP) team (Andrews et al. 2018) using the Atacama Large Millimeter/submillimeter Array (ALMA) who imaged disk substructures, and by scientists using the Spitzer Space Telescope InfraRed Spectrograph (Spitzer-IRS) who provided chemical line fluxes (e.g. Pontoppidan et al. 2010; Carr & Najita 2008). The Spitzer-based chemical studies uncovered potential relationships between the properties of a disk and its molecular emission. In order to build on and discover more of these connections, as well as to better understand these disks and their chemical makeup, data from the James Webb Space Telescope (JWST) is needed. The JWST Disk Infrared Spectral Chemistry (JDISC) team will employ JWST’s Mid-Infrared Instrument Medium Resolution Spectrograph (MIRI-MRS) to capture the spectra for 17 protoplanetary disks; thanks to the instrument's higher spectral resolution, as compared to the Spitzer-IRS, researchers will be able to examine the disk spectra in greater detail. In this paper, we compiled the known properties of 31 protoplanetary disks and calculated line fluxes from previously collected spectra for the following molecules and atoms: H2O, OH, HCN, C2H2, CO2, and Ne II. Using these pre-existing data, we created plots to illustrate the parameter spaces occupied by our protoplanetary disks, thus identifying empty areas to probe with future studies. Additionally, we investigated possible trends between disk chemistry and disk features that might warrant further exploration.

Our research project explored posters created by the Contemporary Physics Education Project, a non-profit organization focused on providing resources for educators that they can use in their classes. The purpose of our project was to gain a better understanding of contemporary physics topics through the exploration of these posters and other online resources, as well as to assess if these resources are suitable for educating a general audience. There are five posters focused on different topics: fundamental particles, fusion, history and fate of the universe, nuclear science, and gravitation. We studied and evaluated the information on the posters, utilized the resources that were available on the organization’s website, and found our own sources for additional information. After each of five research periods, we met as a group to discuss our findings, later creating videos demonstrating our newfound knowledge of each concept.

Through the observation and investigation of each poster, we were able to identify a handful of either key points or points of interest for us regarding each topic. The particle physics poster emphasizes the fact that for every particle type there is a corresponding antiparticle type with identical mass and spin, but opposite charge. The fusion poster taught us about the proton-proton chain and how it is responsible for providing most of the sun’s energy. The history and fate of the universe poster discussed how dark matter was discovered, and explained how scientists have been able to identify the accelerating expansion of the universe. The nuclear science poster highlighted the different phases of nuclear matter as well as the existence of radioactive decay. Lastly, the gravitation poster connected Newtonian gravity to Einstein’s General Theory of Relativity while also providing an explanation of gravitational waves. These details are just a few of the many interesting ideas observed throughout the course of our summer research project.

Radio emissions via magnetic star-planet interactions (SPI) can possibly be observed on various low-mass M dwarf stars with close-in planets. This expectation is based on Jupiter’s aurora: the magnetic interaction between Jupiter and its moon Io leads to energy traveling back to the surface of Jupiter, causing electrons to spiral down to the surface of Jupiter and then emitting radio waves that we can observe from Earth. Based on models of M dwarf star-planet systems, GJ 367 is a candidate for displaying Electron Cyclotron Maser (ECM) emission via star-planet interactions, which should appear as bursts of radio waves at certain orbital positions. I used data from the Australian Telescope Compact Array (ATCA) of GJ 367 to image the visibility data and then create time series of the star’s flux density of its radio waves over time. The time series presents a burst of radio emission, which could be an occurrence of the stellar aurora from GJ 367, but the results are still inconclusive. While the stellar aurora caused by SPI is still not certain for GJ 367, future observations will target GJ 367 further as well as other M dwarfs with the potential of stellar aurora caused by SPI.

Psychological Science

Generalization refers to the transfer of conditioned responding to stimuli that perceptually resemble, but do not exactly match, the initial conditioned stimulus. Previous research has shown the potential for a “top-down” role of the infralimbic cortex (IL) in modulating fear responses. However, an exact role for the IL in fear memory generalization and discrimination is unknown. We used the ArcCre ERT2xEYFP transgenic mouse to visualize “neuronal ensembles” in the IL cortex. The ArcCre ERT2xEYFP transgenic mouse permits the permanent genetic labeling (a green fluorescent protein, GFP) of activated neurons during learning. Following memory retrieval, immunohistochemistry with antibodies against Arc can be used to identify plasticity-related events underlying memory processes. This system allowed for the detection of activity at two different time points during learning (GFP) and memory retrieval (Arc), i.e., a neuronal ensemble. We set out to characterize neuronal ensembles underlying fear generalization and discrimination by first behaviorally characterizing fear memory generalization and discrimination in ArcCre ERT2xEYFP mice. We undercovered a wide spectrum in conditioned freezing to a novel tone stimulus, allowing us to segregate groups into low fear (“discriminators”) and high fear (“generalizers”) groups. Confocal imaging was used to visualize activated neurons and an open source package in ImageJ was used to quantify cell populations. These experiments are ongoing. A chemogenetic approach (DREADDs) was also used to establish a causal link between IL functionality and fear generalization. Following a systemic injection of clozapine N-oxide (a selective DREADDs actuator), mice transfected with an inhibitory DREADD (CaMKII-hM4D(Gi)) or control AAV underwent behavioral testing to investigate how IL inhibition contributes to the conditioned response. These experiments are also ongoing.

Astrocytes are remarkably heterogeneous neuroglial cells essential to synaptic connectivity, synaptic plasticity, and information processing in the nervous system. While previous research suggests that astrocytes are implicated in hippocampal spatial working memory (Newman et al., 2011), their role in the prefrontal cortex, another region associated with working memory, is not as well characterized. Calcium is a critical intracellular secondary messenger leading to gene expression and other functional outputs. Elevated intracellular calcium levels are an effective signaling mechanism within the astrocyte, making Ca2+ waves an appropriate measure of astrocyte activity. Our goal is to explore the association between astrocyte activity in the medial prefrontal cortex (mPFC) and spatial working memory in Long Evans rats by analyzing changes in calcium signals as the rats navigate a spatial working memory task. We inject an astrocyte-specific adeno-associated virus into the rat mPFC that fluoresces in the presence of calcium. We also implant a gradient refractive index lens probe so the waves can be visualized in vivo via a miniature microscope (“miniscope”). Calcium increases are then recorded during a delayed spontaneous alternation task. Before this task, rats either receive an injection of glucose or saline (control). Research has shown that glucose is primarily taken up by astrocytes during periods of high activity and enhances working memory. After the behavioral task, we collect brain slices and stain them for viral expression. Based on histology, we currently hypothesize that increasing the viral load leads to visible viral expression. However, we have been unable to detect astrocyte activity through the miniscope. In the future, we will continue troubleshooting, as we consider many factors that may impede our ability to observe calcium waves through the miniscope, including lens misalignment and blood cells blocking the view from the lens.

Female sex hormones play wide-ranging and crucial roles in brain function. Notably, they affect cognition and confer neuroprotection via receptors on astrocytes. A population of specialized glial cells in the central nervous system (CNS), astrocytes express a plethora of receptors including transporter proteins that facilitate water transport and aid glutamate recycling, the brain’s major excitatory neurotransmitter. The current study seeks to assess the role of estrogen in astrocytic water balance through aquaporin-4 (AQP4) receptors and glutamate cycling through glutamate transporter-1 (GLT-1) during spatial working memory. As estrogen has been found to improve spatial working memory in the hippocampus and prefrontal cortex and decrease striatum-driven behavioral activity, we quantitatively investigated astrocytes expressing AQP4 and GLT-1 in the aforementioned regions to elucidate possible brain area differences. Female rats were ovariectomized bilaterally (OVX) and treated with either 0 (sesame oil), 4.5, or 45 μg/kg dosages of 17β-estradiol 24 and 48 hours prior to behavioral testing. Vaginal smears were collected prior to injections to confirm ovariectomy and prior to behavior to confirm the efficacy of the drug treatments. Spatial working memory was analyzed through a delayed spontaneous alternation task. Consistent with existing literature, ovariectomy impaired delayed spatial working memory (dSWM) with 17β-estradiol rescuing dSWM in a dose-dependent manner. Additionally, while estradiol doses did not have a significant effect on the levels of AQP4 or GLT-1, variation of staining was seen across brain regions with the highest amount of AQP4 staining in the hilus while GLT-1 staining was greatest in the molecular layer of the hippocampus. Clarifying these relationships can illuminate the clinical applications of female sex hormones in health and neurodegenerative disorders.

Astrocytes are cells within the brain that interface with the blood and can have supportive functions such as intake of resources, export of waste in the brain, and support connections (synapses) within the brain necessary for learning and memory. With the use of Designer Receptors Exclusively Activated by Designer Drugs, or DREADDs, we can use compound 21– a new synthetic ligand that activates DREADDs, but unlike the previously used activator, Clozapine-N-oxide, does not metabolize into clozapine– to activate astrocytes in the prefrontal cortex and hippocampus in rats by increasing intracellular calcium. After injecting the experimental group with the DREADDs virus, and the control group with Enhanced Green Fluorescent Protein (EGFP), we perform a 2 delayed alternation task tests 48 hours apart, injecting c21 30 minutes before one test, and the same amount of saline 30 minutes before the other in order to compare their performance with and without activation of DREADDs. With this, we are hoping to see how the activation of astrocytes affects memory by observing behavior, and morphological effects on GFAP. Preliminary results suggest a decrease in performance in females when compound 21 (c21) is introduced, with no significant difference in male performance. We are also observing a difference in entries performed between those who received the 3 mg/kg and 1 mg/kg doses of c21. This study will help us understand the effect of astrocyte activation in working memory.

Research has shown that involvement with the criminal justice system and the presentation of delinquent behaviors during adolescence has been associated with increased rates of early mortality (Stenbacka et al., 2019; Zane et al., 2019). However, most research on this topic has looked at male populations. The present research aims to investigate the relationship between delinquency in adolescent women and premature mortality through a longitudinal follow-up study of approximately 5,500 women from Hathaway’s 1954 sample from various high schools in Minnesota. Based on previous research for male populations, we hypothesize that women who demonstrate significant antisocial behaviors during adolescence will have a shorter lifespan than those who are not involved in delinquent activity. The women took the Minnesota Multiphasic Personality Inventory (MMPI) as 9th graders in 1954 and, since then, the Minnesota Multiphasic Personality Inventory-Adolescent-Restructured Form (MMPI-A-RF) has been developed for personality assessments specifically for adolescents. We will use their responses to the MMPI items and translate them to the MMPI-A-RF scales in order to measure personality traits that may be associated with antisocial behavior, as manifest in delinquency. Additionally, research on adolescent males shows that delinquency tends to be associated with lower levels of cognitive ability, agreeableness, and conscientiousness, and higher levels of neuroticism (Mõttus et al., 2012), as well as with more avoidant personality and socially difficult personality (Barra et al., 2022). We used the Ancestry Library Edition database to ascertain participants’ mortality, using publicly available records to track any name changes, including birth, marriage, divorce, and death records.

Global mental health needs are increasingly disproportionate to the amount of available psychological support. 1 in 5 people are afflicted by a mental health issue, yet nearly 50 percent do not receive treatment. In order to address this gap in care, this study focuses on task-sharing interventions, such as Problem Management Plus (PM+) and Psychological First Aid (PFA). Task-sharing interventions are centered around training non-specialist lay people and community members to deliver psychosocial support. This type of community based support not only bridges the gap between people who need mental health support and the lack of professional resources, but also allows care to be delivered in non-stigmatizing, culturally sensitive methods—as the people who administer the care are part of the same community as those receiving it. PM+ and PFA are brief, basic, non-specialist delivered interventions that exemplify this approach. PFA is a set of skills and actions to help someone in distress, acting as a direct response, while PM+ is a longer process in which the helper and the client meet once a week for 5 weeks. During this research pilot, two PFA trainings were conducted—one with the University of Bahamas North (UBN) and the other with Emergency Department (ED) practitioners at Salem-Spital in Bern, Switzerland. Results from the UBN training indicated that participants demonstrate high to extremely high levels of confidence when caring for someone in crisis following the training. Results from the pre- and post-training surveys with ED personnel were insufficient to draw conclusions about General Self-Efficacy, however, 75 percent of participants demonstrated high to extremely high levels of confidence in using PFA skills to help distressed individuals following the training. The findings of this research suggest that PFA is a cost-effective, low-scale intervention compared to other psychological treatments and is highly effective in high stress settings.

Studies of normative adolescent personality development show that, from middle school to post-high school, levels of Neuroticism (negative emotion and emotional lability) tend to decrease and levels of Extraversion, Openness, and Agreeableness all increase (Klimstra et al., 2009). As high school plays a critical role in adolescent development, we hypothesized that some high school activities may either assist or inhibit personality maturation. For example, high school athletes tend to be less depressed and anxious (lower in Neuroticism) than their peers (Newcome & Boyle, 1995), and it may be that sports participation strengthens students’ emotion regulation capacities, whether through physical activity alone or through teamwork, and also, that students with more effective emotion regulation select into sports. Our study examined high school sports participation and personality change between 9th and 12th grades. We used the subsample of Hathaway’s 1954 statewide sample of Minnesotan adolescents, coding sports participation in the available yearbooks and using items from 9th and 12th grade Minnesota Multiphasic Personality Inventory (MMPI) assessments. The MMPI includes the majority of items in the Clinical, Higher-Order, and Personality Psychopathology Five (PSY-5) scales of the updated MMPI-A-RF (Minnesota Multiphasic Personality Inventory-Adolescent-Restructured Form), designed specifically for adolescents. We also tested whether sports participation affected trajectories of antisocial behavior by compared sports participation and personality change among four groups of students: childhood initiators of delinquency who persisted through adolescence, childhood initiators who remitted in adolescence, adolescent initiators, and non-initiators.

The purpose of this project is to assess the relationship among personality variables, high school extracurricular activities, and women’s longevity. We conducted a follow-up of Hathaway and Monachesi’s 1954 Minnesota sample of adolescents, which included approximately 5,500 women who took the Minnesota Multiphasic Personality Inventory (MMPI) in 9th grade. We ascertained the mortality of individuals from the sample using the Ancestry Library Edition database and assessed participation in high school activities with information from yearbooks, focusing primarily on sports and music activities. Although sports offerings for women were limited prior to Title IX, we hypothesized that women who were involved in sports would live longer than non-athletes due to the well-established link between physical activity and longevity (Loprinzi, 2015), as well evidence that exercise habits formed in adolescence can extend into adulthood (Haynes et al., 2022). The association of personality with both athletic participation (Laborde et al., 2015; Panza et al., 2020), and with mortality (Angst et al., 2013; Lee et al., 2019) led us to consider personality variables as potential mediators of this relationship. Specifically, we examined the constructs of optimism, as assessed by the Malinchoc et al. 1995 Revised Optimism-Pessimism Scale for the MMPI, depression, as measured by the MMPI-A-RF clinical scale 2, Low Positive Emotions, and Sensation-Seeking, measured by the Viken et al. 2004 Sensation-Seeking Scale for the MMPI. Additionally, while there is little evidence of a direct relationship between high school music participation and longevity, some research suggests that involvement in music is associated with cognitive and social benefits (Rogenmoser et al., 2017; Hanna-Plady & McKay, 2011; Welch et al., 2014; Pentikainen et al., 2021), leading us to expect a positive prospective association between high school music participation and longevity.

Some educators believe students reduce effort in response to lenient grades, whereas others believe lenient grades facilitate learning by instilling confidence and enhancing motivation. The current study aims to examine how lenient versus standard grading impacts test performance and motivation. Current college students (N=43) viewed and were subsequently tested on two online Earth Science and Geography lectures. Participants were randomly assigned to a control (standard grades) or experimental condition (lenient grades). The primary dependent variable was participants’ scores on Test 2 following receipt of a standard or lenient grade. Additionally, demographic data, personality measures, instructor evaluations, and test anxiety scores were obtained. Preliminary results indicate a benefit of lenient grades. The group differences in Test 2 performance approached significance, (Mlenient = 8.45, SD = 1.18; Mstandard = 7.76, SD = 1.61; p = 0.114, t(41) = -1.61). Students who received lenient grades performed roughly 10% higher on Test 2 than those who received standard grades. This difference represents a moderate effect size and a substantive difference of a B versus C+ grade. The strength of an experimental approach is that differences between groups can be attributed to grading practices and not to confounding variables due to random assignment. The limitation, however, is the artificiality of an experiment. Grades assigned in an experiment do not have real-life consequences, such as an impact on one’s transcript or GPA. Further, to the extent that grades are used to communicate relative standings to graduate schools and employers, grade inflation may have consequences that extend beyond the content examined in the present study.

As aging occurs, sensitivity to insulin decreases, causing increases in blood glucose and a decrease ATP production. This increases free radicals in the blood and excitotoxicity of cells which leads to cell damage or death and can be reflected in the cognitive decline seen with age. Ketogenic diets bypass glucose/insulin signaling that is compromised in old animals and could be used to enhance cognition across the lifespan. This regime is high in fat and protein and low in carbohydrates, leading to the production of fat-derived ketone bodies that enter the Krebs cycle for ATP production. Gluconeogenesis occurs in parallel with ketosis and maintains blood glucose at low but steady levels. Our study investigates if the ketogenic diet can maintain cognition and the functioning of hippocampal formation astrocytes in aged rats. In collaboration with Joseph McQuail’s lab from the University of South Carolina, we sectioned and performed immunohistochemistry on the cortices of Fisher Brown Norway rats that were male or female, young adult or aged, and on a ketogenic diet (8 weeks) or a normal diet with similar nutrients. We imaged the perirhinal cortex, a hippocampal formation area near the rhinal fissure, and the caudate putamen, an area thought to be maintained with age, to compare the mean optical density and the mean percent area stained of glutamine synthetase (GS) and glial fibrillary acidic protein (GFAP). GS and GFAP are astrocytic markers that represent a functional (GS) and structural (GFAP) link with neurons. Through analyzing preliminary data, we found the two brain areas are significantly different for all analyses. We also found a trend indicating females have higher mean percent area stained for GS than males. Many brains have yet to be sectioned, and with more data, we hypothesize the ketogenic diet will maintain cognition and astrocytic function in aged rats. If our hypothesis is supported, this diet could be utilized to preserve cognitive functioning.