Files and links (1)
PDFOpen Access (License Unspecified), Open Access
Dissertation
Development of a pattern recognition approach for analyzing flow cytometric data
Doctor of Philosophy (Ph.D.), Drexel University
31 Mar 2006
DOI:
https://doi.org/10.17918/etd-843
Abstract
Background: Flow cytometry is used to study the properties of individual cells. Combining cellular markers for viability with cell surfaces marker expression is routinely used to study various cell lineages. Current classification methods use strict thresholds, or "gates", on the fluorescent intensity of these markers. These techniques are subjective in nature and may not fully elucidate phenotypes of interest. Described in this thesis is the development of objective criteria for phenotypic boundary recognition through statistical pattern recognition, and the application of this approach to multiple studies targeted at understanding in vivo erythropoesis, apoptosis, and the role of the hormone erythropoietin (EPO). Methods: Artificial neural networks (ANNs) were trained to recognize subsets of cells with known phenotypes, and to determine decision boundaries separating these subsets based on statistical measures of similarity. The ANNs were trained with samples whose class membership is known with high probability. The method was developed using an in vitro cultured cell system engineered to be dependent on EPO for survival, and a set of parameters correlated to cellular viability. These ANNs were then applied to classify cells by viability and lineage phenotype from several ex vivo studies, and ultimately to analyze late-stage erythroid progenitor viability in response to exogenous recombinant human EPO (rhEPO). Results: The best performing network was a radial basis function multilayer perceptron (RBFP) as measured by multiple performance indices. Classification of the in vivo data confirmed a dependence on EPO for continued survival in the cultured cells. In vivo, treatment with rhEPO increased the fraction of erythroid precursors in the bone marrow. However, there was no change in the number of apoptotic late erythroid precursors and a statistically significant increase in the number of dead erythroid precursors. Further in vivo studies produced no significant differences in erythropoetic state between control and experimental groups in a murine system with defects in the genes that control production of a receptor-ligand combination believed to be critical for the initiation of apoptosis. Conclusions: A statistical pattern recognition approach was demonstrated to provide an objective rationale for setting decision boundaries for classification in cytometric data. Using this approach we have confirmed that rhEPO inhibits apoptosis and cell death in an EPO dependent cell line, but failed to do so in a murine model of anemia. Our results also suggest that rhEPO may not act as a simple anti-apoptotic agent in the bone marrow. Rather, homeostatic mechanisms may regulate the pharmacodynamic response to rhEPO. Further study of this system revealed additional complexities, most notably that there may be redundancy among the receptor-ligand combinations that initiate apoptosis.
Metrics
32 File views/ downloads
24 Record Views
Details
- Title
- Development of a pattern recognition approach for analyzing flow cytometric data
- Creators
- John R. Quinn - DU
- Contributors
- Leonid Hrebien (Advisor) - Drexel University (1970-)
- Awarding Institution
- Drexel University
- Degree Awarded
- Doctor of Philosophy (Ph.D.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Resource Type
- Dissertation
- Language
- English
- Academic Unit
- School of Biomedical Engineering, Science, and Health Systems (1997-2026); Drexel University
- Other Identifier
- 843; 991014631952104721