Laurel Thompson, Utah Valley University
Life Sciences
Breast cancer is divided into subtypes which are defined by their proteomics, histology, and genetic expression profile. Current methods, therefore, are aimed at testing these, and include DNA microarrays, immunohistochemical staining, and proteomic analysis. These methods are effective classifiers, but are not easily transferable to real-time clinical applications, such as the determination of cancerous status during operation or when taking a biopsy. The determination of molecular subtype by other means would be a significant advancement in cancer detection and treatment. We have made some preliminary studies that suggest high-frequency ultrasound may be sensitive to variations among the cancer subtypes as manifest in cell cultures through their cytoskeletal protein structure, which has a distinct spectral signature. The object of this study was to explore the basis for this variation through a combination of experimental and theoretical analysis. We used first-principal modeling methods and compared the model spectra generated from these to spectra obtained in the cell culture lab. Variations in bulk modulus, cell position and size were modeled and combined with experimental spectra in principal component analysis (PCA), and the Euclidean distances between each principal component of the experimental were found as they relate to the theoretical principal components. A graphical method similar to heat maps used for gene expression profiling was then developed to display the relative distances (similarities) between spectra. The program was tested by comparing experimental spectra of three breast cancer cell lines to model spectra. The results indicate the properties and thus molecular subtypes of breast cancer cells could potentially be determined by comparing their measured spectra to model spectra using a feature classification program such as PCA and that these classifying features can be displayed in a convenient graphical representation according to their spectral similarities.