Engineering – UROP Spring Symposium 2021

Research Discipline(s): Engineering

Big Cats and Big Data: modeling the ecology of predator-prey dynamics and intraguild predation in the jaguar (Panthera onca) and ocelot (Leopardus pardalis)

Although intraguild predation and interspecific killing play a major role in structuring ecosystems and food webs, we lack a mechanistic understanding of the complex behaviors and outcomes for coexistence among carnivore species. Such is the focus of this project; we built an agent-based model (ABM) in NetLogo to simulate competition between felids, specifically jaguars (Panthera onca) and ocelots (Leopardus pardalis). We searched the literature to parameterize t important components of the model such as movement, home range, and intraguild killing frequencies. Simulation runs were performed 1000 times for each level of additional arboreal refuge, recording the number of coexistence outcomes over a predetermined number of ticks to represent time. We used generalized linear models (GLMs) to determine the relationships between spatial refugia on the coexistence outcomes of the jaguar and ocelot model. Progress so far indicates that increased arboreal availability leads to more coexistence between the two species despite overlapping home ranges and occupancy. In a broader context, this modeling approach can give researchers predictive power in complex systems and guide management decisions for protected areas in neotropical ecosystems.

Design of cancer-treating proteins

One of the most elusive diseases known to man in terms of treatments, cures, and diversity is cancer. Despite how far the medical field has come, there are still no cures for any type of cancer and treatment options vary drastically in terms of success depending on the type and stage the disease is caught at. When looking to find better treatments or even possible cures for cancer, the most important factor to consider is how proteins produced by cancer cells differ from normally produced proteins and how these proteins interact with one another. In addition knowing things like a cancer protein’s binding sites and what could interfere with and inhibit the function of cancer proteins is critical when coming up with possible treatments and cures. By analyzing these protein-protein interactions and performing a de novo protein design, it is possible and feasible to find proteins that can target cancer-related proteins and slow or stop the spread of the disease in the body.

Design of Novel Protein to Inhibit PD-1/PD-L1 Protein-Protein Interaction for Cancer Immunotherapy

Cancer is the second leading cause of death in the United States and around the world. All cancer is a result of gene mutations, which can cause the formation of abnormally functioning proteins that change a cell’s behavior from normal to cancerous. Contrary to prior cancer treatments focusing on treatments not native to the human body, this research study aims to harness the natural immune response. This project involves the design of a novel protein sequence using a computational protein design program called UniDesign. This novel protein will be designed to inhibit the PD-1/PD-L1 protein-protein interaction, which is responsible for preventing T-cells from destroying other cells. In cancer patients, the new protein could be used to inhibit the PD-1/PD-L1 pathway, which would allow T-cells to destroy cancer cells. Thus, this treatment utilizes the native immune response as a source of cancer therapy. In order to determine the effectiveness of the novel protein at inhibition, the binding affinity of the new protein sequence will be compared to that of the original PD-1/PD-L1 PPI. As such, the protein developed in this project will have the potential to target the PD-1/PD-L1 PPI to ultimately treat cancer in patients.

3d Online Body Shape Model Development

In the automotive industry, there have been an increasing number of fatalities, easily preventable by building safer vehicles. In the medical field, it is almost impossible to thoroughly test accuracy of medical technology without using a physical model, which are oftentimes very expensive. This study involves the development of realistic user-manipulated 3D body models for usage in the medical, automotive, and engineering fields. Along with my mentor and fellow researcher, we have been building a nuanced body model application using Unity to be put out on the apple store and google play store. While results are inconclusive so far, as the application has not been released yet, it hopefully will facilitate the lives of engineers, researchers, and manufacturers working in a number of industries.

3d Online Body Shape Model Development

In the automotive industry, there have been an increasing number of fatalities, easily preventable by building safer vehicles. In the medical field, it is almost impossible to thoroughly test accuracy of medical technology without using a physical model, which are oftentimes very expensive. This study involves the development of realistic user-manipulated 3D body models for usage in the medical, automotive, and engineering fields. Along with my mentor and fellow researcher, we have been building a nuanced body model application using Unity to be put out on the apple store and google play store. While results are inconclusive so far, as the application has not been released yet, it hopefully will facilitate the lives of engineers, researchers, and manufacturers working in a number of industries.

Pollutant Data Prediction: Protecting our Watershed Through Sensing

Watershed health largely influences quality of life, but it is difficult to quantify. Tracking and reducing pollutants is one way to improve watershed health. Total suspended solids (TSS) is an invaluable measurement in determining the amount of pollutants in a body of water. Despite many studies relating TSS to turbidity (water opacity) or doppler backscatter (sound refraction), there is yet to be a way to accurately predict TSS on a long term basis. This study takes previously calculated correlations between suspended solids concentration (SSC) and doppler backscatter using linear regression and conducts a gaussian process on the data to determine a more accurate model. The model created is based on probabilistic computations with a 95% confidence interval. This model will require initial data points from a site before a model can be developed, unlike the previous model, but the confidence interval will decrease in size as more data points are added. Therefore, we will be able to determine a more accurate model for each specific site. The model developed in this study will then be used to predict TSS values for new sites as TSS and SSC are related measurements, and TSS is a more widely used measurement in regulations. This result allows us to continue with further research into the number of true TSS/ SSC samples necessary to determine an accurate model and how the model will react to changes in flow rate implemented by controlled valves.

Eye tracking to determine cognitive load during programming practice

Novice programmers need scaffolded instruction to maximize their ability to learn how to program. Parsons problems are an increasingly popular solution. These problems require learners to place mixed-up code blocks in the correct order to solve a problem. We are conducting think-aloud sessions and a within-subjects experiment to understand the efficiency and cognitive load of solving adaptive Parsons problems versus writing the equivalent (isomorphic) code. We are also investigating the impact of prior programming experience on students’ experiences and changes in students’ self-efficacy. This study will report on cognitive load and self-efficacy ratings before and after the task for the two problem times. We expect students to exhibit greater learning gains on fixing code with errors when solving Parsons problems with distractors than without. We also expect to find a correlation between self-efficacy and cognitive load ratings. The implications of this study are to improve programming learning tools for novice programmers.

Eye tracking to determine cognitive load during programming practice

Novice programmers need scaffolded instruction to maximize their ability to learn how to program. Parsons problems are an increasingly popular solution. These problems require learners to place mixed-up code blocks in the correct order to solve a problem. We are conducting think-aloud sessions and a within-subjects experiment to understand the efficiency and cognitive load of solving adaptive Parsons problems versus writing the equivalent (isomorphic) code. We are also investigating the impact of prior programming experience on students’ experiences and changes in students’ self-efficacy. This study will report on cognitive load and self-efficacy ratings before and after the task for the two problem times. We expect students to exhibit greater learning gains on fixing code with errors when solving Parsons problems with distractors than without. We also expect to find a correlation between self-efficacy and cognitive load ratings. The implications of this study are to improve programming learning tools for novice programmers.

C++ Programming for Data Packaging

The University of Michigan’s Direct Brain Interface Laboratory has a backlog of raw data that needs to be systematically processed and saved into a standardized format for later analysis. This project was to create a new Graphical User Interface (GUI) in MATLAB which would allow any user to simply process and package the raw BCI data. The GUI also needs to take in information gathered from surveys for each session for each participant and save them for later use. Using MATLAB’s design environment App Designer, a new GUI was created that incorporated code from previous data packaging projects. The GUI currently allows the user to load in raw data then process it, and administer and save surveys or load in survey data that was administered elsewhere, then save all of the processed data together in a standardized form. This new GUI will work for the lab’s data collected during the keyboard replacement study, with the ability for data from different protocols to be added in the future.

Species Distribution Modeling of Pseudopipra pipra in the Neotropics

The Neotropics is one of the most species-rich regions in the world (Condon et al., 2008). Its diverse landscape and complex history has allowed for population isolation and subsequent diversification (e.g. Berv et al., 2021). While our knowledge of phylogenetic diversity in the Neotropics has dramatically increased in recent years (e.g. Berv et al., 2021; Tello et al., 2009), our understanding of how species’ have adapted across a myriad of climate regimes in the Neotropics is less well understood (Weir, 2006).

lsa logoum logo