Scholars of Excellence

This research explores the dual role of NRF2, a key regulator of oxidative stress response, in non-small cell lung cancer. It was observed that Nrf2 activation in a genetically engineered mouse model can promote lung tumor initiation in the presence of oncogenic Kras mutation, but excessive Nrf2 activation can inhibit tumor progression by causing metabolic imbalances such as reductive stress, which can impair cancer cell proliferation.

These insights can guide the development of targeted therapies that modulate NRF2 activity depending on the tumor stage and genetic context. Specifically, increasing NRF2 activity in NRF2-hyperactive tumors could drive them to a state that impairs progression. However, since there seems to be an optimal threshold of NRF2 activity, caution is needed when developing NRF2 inhibitors, as reducing NRF2 activity in certain tumor cells could inadvertently permit tumor progression. The study also suggests that NRF2 activators could be useful for chemoprevention in healthy individuals, as Nrf2 activation alone did not initiate tumors in our mouse models. However, long-term human studies are needed to confirm the safety and efficacy of such interventions.

By unveiling the context-dependent effects of NRF2, this research offers promising avenues for personalized cancer therapies that consider the specific genetic and metabolic characteristics of tumors.

Over the past twenty years, scientists have been able to identify special types of electrons in materials as a collective phenomenon. These special materials are known as semimetals, and come in a few different types which hold great potential for applications in modern computing, sensing, and renewable energies. Two problems with the realization of these systems in practical applications is the ability to induce magnetism and systematically control their properties. This dissertation research explored two specific types of semimetal systems, Dirac and Weyl semimetals, and investigated how changing the interfacial and bulk magnetic properties can be used to solve the aforementioned problems.

This dissertation advances both the methodological and theoretical understanding of borderline personality disorder (BPD) by leveraging formal computational methods to explore an alternative model of BPD and its development. This approach offers a powerful framework for generating precise, testable predictions that adequately captures the intricate dynamics underlying the development and maintenance of BPD. The use of formal models in this context is particularly groundbreaking, as previous research has largely overlooked these methods for studying BPD and its development.

By utilizing formal models, this dissertation offers novel insights into the development of BPD, while simultaneously addressing key limitations of existing theoretical frameworks. These findings also have several clinical implications, with the potential to advance current prevention and intervention strategies for BPD. For instance, modeling the complex interactions and risk factors contributing to BPD can shed light on how individuals with varying risk factors and symptoms may respond to different treatments. Understanding these differences may guide clinicians in selecting the most suitable intervention for a patient's unique risk profile and symptomatology, potentially leading to more favorable treatment outcomes.

As the volume and diversity of Internet-of-Things (IoT) data continues to grow, traditional cloud-based processing methods face significant challenges, including latency, bandwidth constraints, and privacy concerns. Our research focuses on employing decision trees (DTs) as an intelligent filtering mechanism on the edge. Preliminary comparisons across four datasets revealed decision trees (DT) are significantly more efficient than multilayer perceptron (MLP), saving 97-98% in area and power, leading to the selection of DT for our proposed architecture for lightweight IoT devices We propose a novel programmable and scalable custom ASIC architecture designed for Decision Tree based ML inference. It incorporates two 8-bit serial in serial out input registers connected to an 8-bit comparator for data processing, the output of the comparator drives the select line of Mux, which selects the respective true and false paths. It has two modes of operation, which can be controlled by asserting mode signals. When the mode is asserted high, the circuit behaves as a leaf node, and we can store classification label values in it, whereas when the mode is not asserted, it behaves as a normal comparator, which decides true and false paths. Our design processes trained datasets for classification through two operational phases. Initially, in the programming phase, attribute values are assigned to respective bit slices along with leaf classification values. In the real-time phase, the inferencing takes place by loading the feature values for comparison at each bit slice, resulting in classification. We have designed and implemented the architecture in the Cadence Virtuoso Layout Design platform with a 0.5um technology node. A 3-level decision tree occupies an area 9,34,241 µm² with 14mW power consumption with a maximum speed of 12.8 MHz.