Project in Psychology
EVALUATION OF PHYSIOLOGICAL AND SUBJECTIVE MANIFESTATIONS OF RETRONASAL OLFACTORY HABITUATION
Anthony Armanious, Judy Gao, Adrija Kundu, Aasha Page, Angelina Resal, Jacqueline Shen, Christopher Shin, Olivia Taylor, Ethan Wang, Clara Yu, Alena Zhang, Xinyi Christine Zhang
Advisor: Dr. Graham A. Cousens
Assistant: Joo Un Lee
Abstract: Despite the long-held notion that humans’ sense of olfaction is relatively primitive, the higher-level processing involved in human olfaction indicates that it is more complex than commonly thought. For example, the human olfactory system undergoes biological learning in the form of habituation, whereby odor intensity perception decreases following prolonged exposure to an olfactory stimulus. Previous literature has explored in great depth the underlying mechanisms of orthonasal olfactory habituation, but the mechanisms of retronasal olfactory —odor perception through the oral cavity—habituation has not been examined as thoroughly. For this reason, our study focuses on retronasal olfaction and the degree to which humans habituate to retronasal olfactory stimuli. We attempt to explore the behavioral and physiological effects of retronasal habitation through participants’ subjective odor intensity ratings in conjunction with their intranasal airflow patterns. Furthermore, we strive to examine the effects of odorant concentrations on habituation as well as the targeted olfactory and trigeminal pathways. Overall, the findings of this study have brought deeper insight into neurological and other physiological mechanisms associated with habituation and retronasal olfaction.
T4 Presentation
Project in Computer Science
APPLICATION OF MACHINE LEARNING FOR ASSESSING MOVEMENT
DISORDERS
Alma Alex, Ashita Birla, Rithvik Bonagiri, Patrick Dugan, Noureldin Elhelw, Emma Fylstra, Jue Gong, Lily Hezrony, Jai Kasera, Saanvi Mehta, Nikhil Mudumbi, and Kayla Shan
Advisor: Dr. Minjoon Kouh
Assistant: Naria Rush
Abstract: Chronic kidney disease (CKD) is a potentially fatal disease that affects a large percentage of the world population. Once CKD has progressed to the end stage, patients must receive treatment to survive, which often comes in the form of renal replacement therapy (RRT). However, predicting when to start RRT can be challenging. A recent study explored the possibility of only using patients’ comorbidities to train ML algorithms to predict whether a patient would need to start RRT within a given period. The purpose of our project was to replicate this study by applying ML techniques to data about the comorbidities of patients with CKD to predict whether they will need RRT within the next 6 months. We used data from Taiwan’s National Health Insurance Database (NHIRD) to train eight machine learning algorithms. These models were then evaluated by their accuracy, sensitivity, specificity, and area under the receiver operating curve (AUC). Four out of the eight algorithms that we tested had AUC greater than 0.70, with XGBoost being the most effective for this dataset. Improvements in computing power, access to more diverse data, and further tailored algorithms will allow ML to be used in hospitals in the future to predict the time of RRT.
T9 Presentation
Project in Physics
QUANTUM COMPUTING
Raghav Akula, Jeanus-Luc Canlapan, Yamato Hara, Emily Harms, Sanskar Jain, Flora Jeng, Sophia Kim, Alex Lin, Alex Noviello, Dhruv Raghuraman, Kevin Song, Brandon Stobie
Advisor: Dr. Daniel Kaplan
Assistant: David Van Dongen
Abstract: Quantum computing is an emerging technology at the forefront of science. We briefly overview the fundamental principles and history of quantum computing before discussing the unique strengths of quantum computers and demonstrating the quantum advantage in optimization problems. We show how quantum computers can effectively group individuals by their relationships. Additionally, we provide evidence of the scalability of quantum algorithms as compared to classical counterparts. We then describe how these computations can be negatively affected by noise and consider how to account for this in future use. Our analysis of noise includes the application of the Central Limit Theorem and the Empirical Rule to calculate broad bounds for future expected error, in the context of circuits consisting primarily of NOT and Hadamard gates, two fundamental operators discussed in detail. We also created an error distribution for an empty quantum circuit, centered around 1.38%, with a standard deviation of 0.395%. Finally, we examine the concepts of quantum entanglement and the No-Cloning Theorem and incorporate the two into quantum teleportation. We demonstrate how these ideas are necessary in quantum computing and in future quantum technologies.
T16 Presentation
Project in Chemistry
COMPUTATIONAL INVESTIGATION OF ANCIENT ALDO-KETO REDUCTASE
STRUCTURE, FUNCTION, AND REACTIVITY
Anmol Bhatia, Isabel Chan, Xiaoyu Cui, Olivia Duarte, Catherine Feng, Kaitlyn Fiore, Harrison Holmes, Matthew Lee, Jonah Perelman, Claudia Sterling, Bethany Mariel Suliguin
Advisor: Dr. Adam Cassano
Assistant: Aline Carla Kruger
Abstract: Aldo-keto reductases (AKRs) are known for their ability to catalyze a wide variety of substrates and have the potential to catalyze reactions important for pharmaceutical and industrial use. Using computational software, the structure and function of AKRs 169, 304, and 308 in Saccharomyces cerevisiae, were studied with a focus on homology, active site binding, and substrate screening to determine candidate molecules for AKR catalysis. Computational docking was executed to predict binding compatibility between the substrates and the AKRs. Docking results that placed substrates near the enzyme cofactor NADPH were used to determine the active site of the proteins. Previous work has found that AKR 169, which corresponds to the protein AKR5F, is capable of detoxifying vanillin to vanillyl, a major component for yeast survival with industrial applications. It was also determined that AKR 304 is homologous to GRE3P and can act as a methylglyoxal reductase. It was also found that AKR 308, which is homologous to Ara1p, is capable of reducing ethyl acetoacetate, ethyl 4-chloroacetoacetate, and 2,3-pentanedione and exhibits enzyme activity not modeled by traditional Michaelis-Menten kinetics. Computational homology tests showed that AKRs are highly similar with present-day yeast strains, and laboratory results displayed AKR 308’s ability to catalyze the reduction of ethyl acetoacetate, ethyl 4-chloroacetoacetate, and 2,3 pentanedione suggesting that AKR 308 performs best with linear substrates.
T17 Presentation
Project in Chemistry
COMPUTATIONAL DRUG DOCKING TO IDENTIFY POSSIBLE THERAPEUTICS THAT WARRANT FURTHER INVESTIGATION FOR ZOONOTIC DISEASES
James Bachman, Zayd Elhedoudy, Andrew Garcia, Keya Gulati, Shruti Mandrekar, Luke Mosca, Phaedra Salerno, Karie Shen, Dylan Shugar, Edward Sun, Anay Tillu, Lucas Wang
Advisor: Dr. David Cincotta
Assistant: Abagail Pedroso
Abstract: Zoonotic diseases are spread from animals to humans, or vice versa. Current zoonotic diseases for which there is no reliable or effective treatment include Ebola virus disease (EVD), rabies, and Zika virus disease (ZVD). To identify possible lead compounds for these diseases, the computational drug docking software SeeSAR was used to simulate interactions between proteins specific to each disease—VP35 and VP40 for the Ebola virus (EBOV), the rabies virus (RABV) glycoprotein, and the Zika virus (ZIKV) envelope protein—and ligands from a list of 3,035 FDA-approved drugs. Drugs from this list were initially screened against each protein using 20 poses. Drugs with at least one pose that had an estimated lower bound dissociation constant (K D ) below 100 nM, indicating favorable binding affinity, were screened again using 200 poses. Furthermore, chemical modifications were performed on DOTAP chloride, a drug possibly effective against EBOV, which led to sizable improvements in the binding affinity. Some possible therapeutic drug candidates are as follows: modified DOTAP chloride for EBOV VP35; gefarnate, salmeterol, and deferoxamine mesylate for EBOV VP40; olivetol, aleuritic acid, and bimatoprost for the RABV glycoprotein; and deferoxamine mesylate, tirofiban hydrochloride, and travoprost for the ZIKV envelope protein. The criteria for recommendation include possible side effects, number of poses under 100 nM, and K D values. This study’s results display the candidate drugs’ potential to be investigated further as novel lead compounds in clinical trials for EVD, rabies virus, and ZVD.
T18 Presentation