Session C: 1:45PM – 3:15PM
SESSION C (1:45PM – 3:15PM)
Location: Conference Room, Sill Center
Don’t Fear the AI: A Systematic Review of Machine Learning for Prostate Cancer Detection in Pathology
Alexander Gibson, Brigham Young University
Faculty Mentor Dennis Della Corte, Brigham Young University
SESSION C 1:45-2:00PM
Health and Medicine
Prostate cancer is the second-most diagnosed cancer in men worldwide. It is also the fifth leading cause of death in men. When a patient is referred to a pathologist due to the presence or suspicion of prostate cancer, a pathologist may take a needle core biopsy, which produces six to twelve slides of prostate tissue. The pathologist then stains the tissue (usually using a hematoxylin and eosin stain) and analyzes the tissue samples using a Whole Slide Scanner. The scanner digitizes physical slides into high resolution giga-pixel images. As the shortage of pathologists continues, remaining pathologists are left to analyze more samples than before. Pathologists also face significant inter-observer variability. Thus, the use of artificial intelligence (AI) can drastically increase the productivity and impact of a single pathologist. Researchers across the globe are inventing new machine learning algorithms to detect and classify prostate tissue. This literature review explains how artificial intelligence is being used to detect prostate cancer. After reading over 400 papers, we narrowed our pool of papers down to 142 relevant papers. We classified all machine learning results and methods into 11 different categories (see figure in uploaded PDF). Papers reported algorithm performance according to many different metrics, but we determined that Kappa values, Area under the Curve values, and Accuracy values are the most meaningful metrics. Some papers used a binary classification method to classify tissue, and some papers used a multi-class method. All relevant algorithms were categorized according to their methods and performance. We conclude that pathologists will not be replaced by machine learning algorithms; however, AI can enhance the performance and efficiency of pathologists. An increase in pathologists’ effectiveness could improve access to healthcare in underdeveloped areas and emergent nations.
Chimeric Autoantigen Receptor (CAAR) T cells as a Novel Immunotherapy for Autoreactive B Cells in Graves’ Disease
Mackenzie Hansen, Brigham Young University
Faculty Mentor Kim O’Neill, Brigham Young University
SESSION C 2:05-2:20PM
Health and Medicine
Graves’ Disease is the fourth most common autoimmune disease in the US. The main cause of Graves’ Disease is the overstimulation of the thyroid gland by thyroid stimulating hormone receptor (TSHR) specific antibodies produced by autoreactive B cells. Current therapies for Graves’ Disease include antithyroid drugs, radioiodine therapy, and surgery, but these do not address the underlying mechanism of the autoimmune response. Our aim is to generate a targeted method to attack the disease using engineered chimeric autoantigen receptor (CAAR) T cells. Our CAAR T cells contain varying epitopes of TSHR that autoreactive B cells will recognize, bind to, and activate. The activated CAAR T cell will then kill the autoreactive B cell. We will compare our candidate CAAR T cells to see which epitope expresses and binds most effectively. We will also perform cytotoxicity assays to measure the targeting and killing ability of our CAAR construct against B cells from Graves’ Disease patients. The use of CAAR T cells specifically targeting autoreactive B cells would open a new avenue of treatment for Graves’ Disease and potentially other autoimmune diseases.
Intact endothelial cell autophagy preserves outcomes of acute ischemic stroke in mice
Milo Light, University of Utah
Faculty Mentor John Symons, University of Utah
SESSION C 2:25-2:40PM
Health and Medicine
Approved treatments for acute ischemic stroke (AIS) include thrombolysis (clot dissolution) and thrombectomy (clot removal). Most patients are ineligible for these procedures because they must be initiated within 4.5h (thrombolysis) or 24h (thrombectomy) of AIS symptom onset. New targets for intervention are needed. Here we evaluate the contribution from endothelial cell (EC) autophagy to outcomes of AIS. AIS creates a nutrient stress and activates EC autophagy. Heightened EC autophagy in response to AIS helps to : (i) identify, tether, and shuttle damaged proteins to the lysosome for degradation and recycling; (ii) generate ATP from recycled macromolecules; and (iii) preserve arterial function by enabling EC nitric oxide production. First we hypothesized that depleting EC autophagy worsens outcomes of AIS. Adult mice with intact autophagy (ATG3 WT ) or depletion of autophagy regulated gene 3 (Atg3) specifically in ECs (ATG3 EC-/- ) were challenged with 60-min middle cerebral artery occlusion followed by 23 h reperfusion. By design, AIS increased (p<0.05) EC autophagy in brains from ATG3 WT but not ATG3 EC-/- mice. Infarct volume was larger, and neurobehavioral and physical deficits were more severe (all p<0.05), in ATG3 EC-/- vs. ATG3 WT mice (n=7 per group). Second, we hypothesized that amplifying autophagy improves outcomes of AIS. For 3-weeks adult mice consumed standard chow that was (rapa) or was not (control) supplemented with the mammalian target of rapamycin complex 1 (mTORC1) inhibitor rapamycin, a potent activator of autophagy. As evidence of mTORC1 inhibition, fasting-induced p-s6K : s6 was greater (p<0.05) in liver segments from rapa vs. control mice (n=4 per group). AIS-induced outcomes concerning infarct volume, and neurobehavioral and physical performance, were superior in rapa vs. control mice (n=8 per group). These results indicate EC autophagy depletion worsens, whereas EC autophagy activation mitigates, outcomes of AIS. Our ongoing studies are targeting EC metabolism to improve outcomes of AIS.
Identification and Analysis of Compounds To Improve CTNNB1-mutated Hepatocellular Carcinoma (HCC) in Transgenic Zebrafish
Audrey Su, University of Utah
Faculty Mentor Kimberley Evason, University of Utah
SESSION C 2:45-3:00PM
Health and Medicine
Hepatocellular carcinoma (HCC) is the most common primary liver cancer with a 5-year survival rate of <20%, and it was the sixth leading cause of cancer mortality in 2020. One of the most commonly mutated genes in HCC is β-catenin (CTNNB1), accounting for ~20-40% of patients with HCC and leading to constitutively active Wnt signaling. Due to the necessity of Wnt signaling in healthy liver tissue, there are few precision medicine-based treatments for patients with either CTNNB1-mutated HCC or HCC as a whole. Instead of targeting specific mutated protein(s), this research focuses on analyzing dysregulated pathways downstream that are susceptible to treatment. Our aim is to characterize the effects of mutated CTNNB1 on lipid metabolism and tumorigenesis. To understand the role of lipid metabolism in HCC, we screened 194 metabolic/protease-related compounds using our lab’s transgenic CTNNB1 zebrafish model of HCC. Wild-type and transgenic larvae were exposed to either DMSO control or experimental compound from 3 to 6 days post fertilization (dpf), and liver size was recorded at 6 dpf. Compounds that decreased liver size in transgenic zebrafish were considered positive hits. We confirmation-tested 10 compounds with additional doses and greater sample sizes. Our results showed that FAAH-IN2 and pitavastatin decreased larval liver size in transgenic zebrafish, while GSK1940029 increased liver size. To further characterize these compounds’ effects, we used oil red-o staining to analyze changes in lipid storage. Our work successfully identified key compounds that significantly affect larval liver size in transgenic zebrafish. These findings could lead to better treatments for both CTNNB1-mutated HCC and HCC as a whole. Future work includes studying differences in hepatocyte proliferation and analyzing alterations in immune cell migration in zebrafish treated with FAAH-IN2, pitavastatin, and GSK1940029, as well as continued confirmation testing in remaining compounds.