Date of Award
Master of Science
Hospital-acquired infections (HAIs) have high mortality rates around the world and are a challenge to medical science due to rapid mutation rates in their pathogens. A new methodology is proposed to identify bacterial species causing HAIs based on sets of universal biomarkers for next-generation microarray designs (i.e., nxh chips), rather than a priori selections of biomarkers. This method allows arbitrary organisms to be classified based on readouts of their DNA sequences, including whole genomes. The underlying models are based on the biochemistry of DNA, unlike traditional edit-distance based alignments. Furthermore, the methodology is fairly robust to genetic mutations, which are likely to reduce accuracy. Standard machine learning methods (neural networks, self-organizing maps, and random forests) produce results to identify HAIs on nxh chips that are very competitive, if not superior, to current standards in the field. The potential feasibility of translating these techniques to a clinical test is also discussed.
dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.
Pham, Duy Tran, "Genomic Methods for Bacterial Infection Identification" (2018). Electronic Theses and Dissertations. 1799.