Presenter: Tyson Hoyt, College of Science, Biology
Authors: Tyson Hoyt, Dr. Vern Hart
Faculty Advisor: Vern Hart, College of Science, Physics
Institution: Utah Valley University
Over the past two years, our research team has successfully developed an imaging technique capable of differentiating cancer cell types using the diffraction patterns from a near-infrared diode laser. We are currently in the process of training an artificially intelligent neural network to distinguish between various cancer cell concentrations (e.g., 5% cancer vs 6% cancer in a background of healthy cells). High precision in these measurements could allow the algorithm to identify cancer cells in small quantities (i.e., 1% cancer vs 0% cancer), leading to earlier detection of micro-cancers and circulating tumor cells (CTCs). To assist with the labeling of desmoplastic cerebellar medulloblastoma cells (DAOY) and Uppsala malignant glioma cells (U87), I will be transfecting one of each cell line with a green fluorescent (jellyfish) protein. When placed under an ultraviolet (UV) light source, the transfected cells will emit a soft green light that will be used to visibly distinguish each cell line. This will make differentiating cell types significantly easier and will allow for the use of auto-segmentation algorithms to count individual cells in microscopy images. My primary role in this project will be the development of transfection protocols using a plasmid to infect cell genomes without altering their shape, function, or resulting optical diffraction pattern. Some cells will not absorb the plasmid, so we are using an antibiotic to kill only the non-transfected cells. Over the past few months, I have optimized the antibiotic concertation needed to kill the remaining cells at the desired rate (~50% of the line within 5 days). This funding will allow me to apply the developed antibiotic technique to cell lines being grown by other members of our research group, to assist in the imaging of cell lines, the acquisition of training data, and future improvements to the machine learning model.